CN110288605A - Cell image segmentation method and device - Google Patents
Cell image segmentation method and device Download PDFInfo
- Publication number
- CN110288605A CN110288605A CN201910504393.2A CN201910504393A CN110288605A CN 110288605 A CN110288605 A CN 110288605A CN 201910504393 A CN201910504393 A CN 201910504393A CN 110288605 A CN110288605 A CN 110288605A
- Authority
- CN
- China
- Prior art keywords
- cell
- image
- segmentation
- enhancing
- cell image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
- G06V20/695—Preprocessing, e.g. image segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
- G06V20/698—Matching; Classification
-
- 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
- G06T2207/10056—Microscopic image
-
- 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
- G06T2207/30024—Cell structures in vitro; Tissue sections in vitro
Abstract
The embodiment of the present invention provides a kind of cell image segmentation method and device, the method comprise the steps that obtaining cell image to be split；Cell image to be split is input in cell segmentation model, the cell segmentation result of cell segmentation model output is obtained；Wherein, cell segmentation model is to be trained based on sample cell image and the corresponding true tag image of sample cell image to enhancing U-Net network；Enhancing U-Net network is the increase network number of plies in initial U-Net network, and BN layers of neural network is added in a network.Method and apparatus provided in an embodiment of the present invention, in the segmentation effect for improving overlapping cell and adhesion cells based on initial U-Net network, while optimizing cell segmentation accuracy rate, based on the training of BN algorithm acceleration model, Optimized model performance.Cell image to be split is input to the cell segmentation model that thus training obtains, quick, accurate, easy cell segmentation can be realized.
Description
Technical field
The present invention relates to technical field of computer vision more particularly to a kind of cell image segmentation method and devices.
Background technique
Cell segmentation is most basic in medical image analysis processing while being also mostly important one of content, is cell
One basic premise of follow-up works such as identified, counted.The segmentation of cell image to quantitative analysis and processing cellular informatics,
It studies cytometaplasia and realizes that cell is micro- or the three-dimensional reconstruction of ultrastructure etc. has irreplaceable realistic meaning.
Traditional cell image segmentation method is roughly divided into two classes: the dividing method based on region and point based on edge
Segmentation method.Wherein, the basic principle of the dividing method based on region is by the way that the adjacent area with similar features is classified as one
Class is divided to realize, representative dividing method has threshold method, region-growing method and clustering procedure etc..Based on edge
Dividing method is representative generally by there is the place of mutation to be split as edge gray level or structure
Method has method of differential operator, modelling etc..
However, above-mentioned cell segmentation method is poor for the segmentation effect of overlapping cell and adhesion cells, step is complicated, operation
Amount is big, and accuracy rate is low.
Summary of the invention
The embodiment of the present invention provides a kind of cell image segmentation method and device, to solve existing Methods of Segmentation On Cell Images
Problem poor for the segmentation effect of overlapping cell and adhesion cells, step is complicated, operand is big, accuracy rate is low.
In a first aspect, the embodiment of the present invention provides a kind of cell image segmentation method, comprising:
Obtain cell image to be split；
The cell image to be split is input in cell segmentation model, the thin of the cell segmentation model output is obtained
Born of the same parents' segmentation result；Wherein, the cell segmentation model is corresponding true based on sample cell image and the sample cell image
Real label image is trained enhancing U-Net network；The enhancing U-Net network is in initial U-Net network
Increase the network number of plies, and BN layers of neural network is added in a network.
Second aspect, the embodiment of the present invention provide a kind of Methods of Segmentation On Cell Images device, comprising:
Image acquisition unit, for obtaining cell image to be split；
Image segmentation unit obtains described thin for the cell image to be split to be input in cell segmentation model
The cell segmentation result of born of the same parents' parted pattern output；Wherein, the cell segmentation model is based on sample cell image and the sample
The corresponding true tag image of this cell image is trained enhancing U-Net network；The enhancing U-Net network
It is the increase network number of plies in initial U-Net network, and BN layers of neural network is added in a network.
The third aspect, the embodiment of the present invention provide a kind of electronic equipment, including processor, communication interface, memory and total
Line, wherein processor, communication interface, memory complete mutual communication by bus, and processor can call in memory
Logical order, to execute as provided by first aspect the step of method.
Fourth aspect, the embodiment of the present invention provide a kind of non-transient computer readable storage medium, are stored thereon with calculating
Machine program is realized as provided by first aspect when the computer program is executed by processor the step of method.
A kind of cell image segmentation method provided in an embodiment of the present invention and device, by increasing in initial U-Net network
The network number of plies is added, and the neural network that BN layers are added in a network is trained to obtain the cell segmentation mould for cell segmentation
Type optimizes the same of cell segmentation accuracy rate in the segmentation effect for improving overlapping cell and adhesion cells based on initial U-Net network
When, based on the training of BN algorithm acceleration model, Optimized model performance.By cell image to be split be input to thus training obtain it is thin
Quick, accurate, easy cell segmentation can be realized in born of the same parents' parted pattern.
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 this hair
Bright some embodiments for those of ordinary skill in the art without creative efforts, can be with root
Other attached drawings are obtained according to these attached drawings.
Fig. 1 is the flow diagram of cell image segmentation method provided in an embodiment of the present invention；
Fig. 2 be another embodiment of the present invention provides cell image segmentation method flow diagram；
Fig. 3 is cell image to be split provided in an embodiment of the present invention；
Fig. 4 is cell segmentation result schematic diagram provided in an embodiment of the present invention；
Fig. 5 is the structural schematic diagram of Methods of Segmentation On Cell Images device provided in an embodiment of the present invention；
Fig. 6 is the structural schematic diagram of electronic equipment provided in an embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without creative efforts, shall fall within the protection scope of the present invention.
For segmentation effect of the existing cell segmentation method for overlapping cell and adhesion cells is poor, step is complicated, fortune
The problem that calculation amount is big and accuracy rate is low, the embodiment of the present invention provide a kind of cell image segmentation method.Fig. 1 is the embodiment of the present invention
The flow diagram of the cell image segmentation method of offer, as shown in Figure 1, this method comprises:
Step 110, cell image to be split is obtained.
Herein, the cell image that cell image to be split needs to be split.
Step 120, cell image to be split is input in cell segmentation model, obtains the thin of cell segmentation model output
Born of the same parents' segmentation result；Wherein, cell segmentation model is based on sample cell image and the corresponding true tag figure of sample cell image
Picture is trained enhancing U-Net network；Enhancing U-Net network is to increase network layer in initial U-Net network
Number, and BN layers of neural network is added in a network.
Specifically, for being split to the cell image to be split of input, cellulation segmentation result is simultaneously exported.This
Place, cell segmentation is the result is that the result that cell segmentation model is split cell image to be split.
Herein, cell segmentation model is to be based on sample cell image and the corresponding true tag image of sample cell image,
Enhancing U-Net network is trained.Sample cell image and true tag image correspond, true tag image
It is to be split to sample cell image.
Enhancing U-Net network is combined with the U-Net network of BN (Batch Normalization) algorithm, i.e., initial
On the basis of U-Net network, the network number of plies is expanded, and hidden by the part that BN algorithm standard handles initial U-Net network
The input data of layer, i.e., add one layer BN layers before the part hidden layer of initial U-Net network.
Wherein, the conventional U-Net network that initial U-Net network is chosen in advance.U-Net network is convolutional neural networks
A kind of deformation, mainly consists of two parts: constricted path (contracting path) and extensions path (expanding
path).Constricted path is primarily used to capture the contextual information (context information) in picture, and phase therewith
Symmetrical extensions path is then to carry out precise positioning (localization) to the part split required in picture.
U-Net network is to realize accurately to position, and the high pixel characteristic extracted in constricted path can be in a liter sampling
(upsampling) it is combined during with new characteristic pattern (feature map), is adopted with retaining front drop to the greatest extent
The some important characteristic informations of sample (downsampling) process；And in order to realize the efficient operation of network structure, U-Net net
Full articulamentum is eliminated in network, largely to reduce the parameter for needing training.In addition, having benefited from special U-shaped structure, U-
Net network can be very good to retain all information in picture.
When in order to alleviate U-Net network training, the problem of data distribution of network middle layer changes, the present invention is implemented
Example handles the input data of the part hidden layer of initial U-Net network using BN algorithm standard, and data input by treated
Extremely in next layer, to improve the speed and constringency performance of network training.
Method provided in an embodiment of the present invention, by increasing the network number of plies in initial U-Net network, and in a network
BN layers of neural network is added to be trained to obtain the cell segmentation model for cell segmentation, is being based on initial U-Net network
The segmentation effect of overlapping cell and adhesion cells is improved, while optimizing cell segmentation accuracy rate, is based on BN algorithm acceleration model
Training, Optimized model performance.Cell image to be split is input to the cell segmentation model that thus training obtains, can be realized fast
Speed, accurate, easy cell segmentation.
Based on the above embodiment, in this method, before step 120 further include: step 100, based on sample cell image and
The corresponding true tag image of sample cell image is trained enhancing U-Net network.
Cell segmentation model can specifically be trained in the following way and be obtained: firstly, collecting great amount of samples cell image and sample
The corresponding true tag image of this cell image；Wherein, sample cell image and true tag image correspond, true tag
Image can be that segmentation obtains by hand by expert.Immediately based on sample corpus labeling task and sample annotation results to increasing
It has added BN layers of initial U-Net network before the network number of plies and part hidden layer in a network to be trained, to obtain thin
Born of the same parents' parted pattern.The embodiment of the present invention does not make specific limit to the scale of initial U-Net network.
Based on any of the above-described embodiment, in this method, before step 100 further include: obtain sample cell image and sample
The corresponding true tag image of cell image: to the data set mark of any data set cell image and the data set cell image
Label image merges, and obtains data set and merges image；To data acquisition system and image carries out data enhancing, obtains data set enhancing
Image；Data enhancing includes at least one of geometric transformation, Stochastic Elasticity deformation, random scaling, shear transformation counterclockwise；It will
Data set enhances image and carries out image separation, obtains the true tag image of sample cell image and sample cell image.
Specifically, the training set for Methods of Segmentation On Cell Images training is compared with the training set in remaining field, procurement cost
It is no matter all bigger in time or in the consumption of resource.In order to solve the problems, such as that training set is too small, need for having obtained
Data set cell image and data set label image expanded.
Herein, data set cell image is the cell image obtained from presently disclosed data set, data set label figure
Segmentation result as being the data set cell image.
For any data set cell image and its data set label image, first to any data set cell image and its
Data set label image merges, and obtains image, that is, data set after merging and merges image.Data set merges both to be wrapped in image
Feature containing data set cell image also includes the feature of data set label image.Immediately to data acquisition system and image into
The enhancing of row data obtains the enhanced data set of data and merges image, i.e. data set enhances image.After data set enhances image,
Data set cell image and data set label image comprising passing through data enhancing, and data set cell image and data set label
The data enhancement operations of image experience are completely the same.After completing data enhancing, image separation is carried out to data set enhancing image,
Obtain data set enhancing image separation after cell image and label image, and by after separation cell image and label image make
For sample cell image and true tag image.That is, being carried out for any data set cell image and its Segmentation of Data Set result
Merging, data enhancing and separation, can be obtained new sample cell image and true tag image, and then realize the expansion of training set
Increase.
Based on any of the above-described embodiment, in this method, before step 100 further include: carry out normalizing to sample cell image
Change processing.
Specifically, it before great amount of samples cell image is applied to cell segmentation model training, needs to each sample
Cell image carries out 0-1 normalized, in order to acceleration model convergence.
Based on any of the above-described embodiment, in this method, before step 100 further include: be based on Gaussian Profile initial method
To initializing.
Specifically, it before training cell segmentation model, needs to initialize enhancing U-Net network parameter.Gauss
The Gaussian Profile that distribution (Gaussian) initialization needs to be generated according to the mean value and standard deviation of preset Gaussian function into
The configuration of row parameter initialization.In method provided in an embodiment of the present invention, parameter is by Gaussian ProfileIt generates, wherein n is
The fan-in of weight tensor.In addition, iteration round epoch=30, batch sample batchsize=4, learning rate η=1e-4.
Based on any of the above-described embodiment, in this method, step 100 is specifically included:
Step 101, sample cell image and the corresponding true tag image of sample cell image are input to enhancing U-Net
Network obtains the training cell segmentation result of enhancing U-Net network output.
Herein, training cell segmentation is the result is that enhance U-Net network to the sample cell image of input in training process
The segmentation result being split.
Step 102, true tag image and training cell segmentation result are input to logarithm loss function, obtain loss letter
Numerical value.
Herein, logarithm loss function is binary_cross_entropy, and binary_cross_entropy is two classification
Cross entropy, two classification herein refer to whether be cell edges.
Step 103, it is based on loss function value, parameter regulation is carried out to enhancing U-Net network by Adam optimization algorithm.
Adam is a kind of first-order optimization method that can substitute traditional stochastic gradient descent process, it can be based on training data
Iteratively update neural network weight.Loss function value is minimized using Adam optimization algorithm, enhances U-Net network weight to update
Weight, until the number of iterations is more than pre-set iteration round epoch.
Based on any of the above-described embodiment, in this method, before step 120 further include: be input to test cell image carefully
In born of the same parents' parted pattern, the test cell segmentation result of cell segmentation model output is obtained；If the corresponding benchmark of test cell image
Cell segmentation result is different from test cell segmentation result, then thin based on test cell image and the training of benchmark cell segmentation result
Born of the same parents' parted pattern.
Specifically, it before the segmentation that application cell parted pattern carries out cell image to be split, needs to trained
Cell segmentation model carries out test verifying.Herein, test cell image is for carrying out test verifying to cell segmentation model
Cell image, benchmark cell segmentation the result is that first pass through the segmentation result that expert is split test cell image in advance.
Test cell segmentation result is the segmentation result of cell segmentation model output.Divide in benchmark cell segmentation result and test cell
As a result when different, based on test cell image and benchmark cell segmentation result training cell segmentation model, to advanced optimize mould
Type performance.
Based on any of the above-described embodiment, Fig. 2 be another embodiment of the present invention provides cell image segmentation method process
Schematic diagram, as shown in Fig. 2, this method comprises:
Step 210, the public data collection that ISBI2012 contest organizing committee provides is obtained.Microscopic cells in the data set
Image sources are in the abdominal nerve cell of the first instar larvae of drosophila.Microscopic cell images, that is, data set cytological map in the data set
Picture.Using 26 width images in the data set as the sample cell image in training set, and for above-mentioned 26 width image respectively into
Row data amplification, to expand the scale of training set.Herein, data amplification needs to meet: (1) expanding resulting sample cell image
In original data set cell image independent same distribution or approximate independent same distribution；(2) resulting sample cell image is expanded to protect
Stay the important feature of original image.
For any data set cell image, data amplification step is as follows:
Any data set cell image data set label image corresponding with the data set cell image is merged, is obtained
Take data acquisition system and image；To data acquisition system and image carries out data enhancing, obtains data set enhancing image；Data set is enhanced
Image carries out image separation, obtains sample cell image and the corresponding true tag image of sample cell image.
Step 220, each sample cell image in resulting training set is expanded to step 210 data to pre-process,
Each sample cell image is converted into the identical Output matrix of size by 0-1 normalized.
Step 230, building enhancing U-Net network.Enhancing U-Net network is to increase network layer in initial U-Net network
Number, and BN layers of neural network is added in a network.After the building of completion, based on Gaussian Profile initial method to enhancing U-
Net network is initialized.
Step 240, it is based on sample cell image and the corresponding true tag image of sample cell image, to enhancing U-Net
Network is trained.Herein, training loss function used is logarithm loss function binary_cross_entropy, is used
Adam optimization algorithm minimizes loss function, updates Model Weight, and the enhancing U-Net network that training is completed is as cell segmentation
Model.
Step 250, cell image to be split as shown in Figure 3 is input to cell segmentation model, obtains cell segmentation mould
The cell segmentation result as shown in Figure 4 of type output.
Method provided in an embodiment of the present invention, by increasing the network number of plies in initial U-Net network, and in a network
BN layers of neural network is added to be trained to obtain the cell segmentation model for cell segmentation, is being based on initial U-Net network
The segmentation effect of overlapping cell and adhesion cells is improved, while optimizing cell segmentation accuracy rate, is based on BN algorithm acceleration model
Training, Optimized model performance.Cell image to be split is input to the cell segmentation model that thus training obtains, can be realized fast
Speed, accurate, easy cell segmentation.
Based on any of the above-described embodiment, Fig. 5 is the structural representation of Methods of Segmentation On Cell Images device provided in an embodiment of the present invention
Figure, as shown in figure 5, the device includes image acquisition unit 510 and image segmentation unit 520；
Wherein, image acquisition unit 510 is for obtaining cell image to be split；
Image segmentation unit 520 is for the cell image to be split to be input in cell segmentation model, described in acquisition
The cell segmentation result of cell segmentation model output；Wherein, the cell segmentation model is based on sample cell image and described
The corresponding true tag image of sample cell image is trained enhancing U-Net network；The enhancing U-Net net
Network is to increase the network number of plies in initial U-Net network, and BN layers of neural network is added in a network.
Device provided in an embodiment of the present invention, by increasing the network number of plies in initial U-Net network, and in a network
BN layers of neural network is added to be trained to obtain the cell segmentation model for cell segmentation, is being based on initial U-Net network
The segmentation effect of overlapping cell and adhesion cells is improved, while optimizing cell segmentation accuracy rate, is based on BN algorithm acceleration model
Training, Optimized model performance.Cell image to be split is input to the cell segmentation model that thus training obtains, can be realized fast
Speed, accurate, easy cell segmentation.
Based on any of the above-described embodiment, which further includes training unit；
Training unit is used to be based on the sample cell image and the corresponding true tag image of the sample cell image,
The enhancing U-Net network is trained.
Based on any of the above-described embodiment, which further includes data amplification unit；Data amplification unit is used for:
Any data set cell image and the corresponding data set label image of any data set cell image are carried out
Merge, obtains data set and merge image；
Image is merged to the data set and carries out data enhancing, obtains data set enhancing image；The data enhance
At least one of geometric transformation, Stochastic Elasticity deformation, random scaling, shear transformation counterclockwise；
Data set enhancing image is subjected to image separation, obtains the sample cell image and the sample cytological map
As corresponding true tag image.
Based on any of the above-described embodiment, which further includes normalization unit；
Normalization unit is for being normalized the sample cell image.
Based on any of the above-described embodiment, which further includes initialization unit；
Initialization unit is used to initialize the enhancing U-Net network based on Gaussian Profile initial method.
Based on any of the above-described embodiment, in the device, training unit is specifically used for:
The sample cell image is input to the enhancing U-Net network, obtains the enhancing U-Net network output
Training cell segmentation result；
The true tag image and the trained cell segmentation result are input to logarithm loss function, obtain loss letter
Numerical value；
Based on the loss function value, parameter regulation is carried out to the enhancing U-Net network by Adam optimization algorithm.
Based on any of the above-described embodiment, which further includes test cell；Test cell is used for:
Test cell image is input in the cell segmentation model, the test of the cell segmentation model output is obtained
Cell segmentation result；
If the corresponding benchmark cell segmentation result of the test cell image is different from the test cell segmentation result,
Based on the test cell image and the benchmark cell segmentation result training cell segmentation model.
Fig. 6 is the entity structure schematic diagram of electronic equipment provided in an embodiment of the present invention, as shown in fig. 6, the electronic equipment
It may include: processor (processor) 601,602, memory communication interface (Communications Interface)
(memory) 603 and communication bus 604, wherein processor 601, communication interface 602, memory 603 pass through communication bus 604
Complete mutual communication.Processor 601 can call the meter that is stored on memory 603 and can run on processor 601
Calculation machine program, to execute the cell image segmentation method of the various embodiments described above offer, for example, obtain cytological map to be split
Picture；The cell image to be split is input in cell segmentation model, the cell point of the cell segmentation model output is obtained
Cut result；Wherein, the cell segmentation model is based on sample cell image and the corresponding true mark of the sample cell image
Image is signed, enhancing U-Net network is trained；The enhancing U-Net network is increased in initial U-Net network
The network number of plies, and BN layers of neural network is added in a network.
In addition, the logical order in above-mentioned memory 603 can be realized by way of SFU software functional unit and conduct
Independent product when selling or using, can store in a computer readable storage medium.Based on this understanding, originally
The technical solution of the inventive embodiments substantially part of the part that contributes to existing technology or the technical solution in other words
It can be embodied in the form of software products, which is stored in a storage medium, including several fingers
It enables and using so that a computer equipment (can be personal computer, server or the network equipment etc.) executes the present invention respectively
The all or part of the steps of a embodiment the method.And storage medium above-mentioned includes: USB flash disk, mobile hard disk, read-only memory
(ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic or disk
Etc. the various media that can store program code.
The embodiment of the present invention also provides a kind of non-transient computer readable storage medium, is stored thereon with computer program,
The computer program is implemented to carry out the cell image segmentation method of the various embodiments described above offer when being executed by processor, such as wraps
It includes: obtaining cell image to be split；The cell image to be split is input in cell segmentation model, the cell point is obtained
Cut the cell segmentation result of model output；Wherein, the cell segmentation model is thin based on sample cell image and the sample
The corresponding true tag image of born of the same parents' image is trained enhancing U-Net network；The enhancing U-Net network be
Increase the network number of plies in initial U-Net network, and BN layers of neural network is added in a network.
System embodiment described above is only schematical, wherein described, unit can as illustrated by the separation member
It is physically separated with being or may not be, component shown as a unit may or may not be physics list
Member, it can it is in one place, or may be distributed over multiple network units.It can be selected according to the actual needs
In some or all of the modules achieve the purpose of the solution of this embodiment.Those of ordinary skill in the art are not paying creativeness
Labour in the case where, it can understand and implement.
Through the above description of the embodiments, those skilled in the art can be understood that each embodiment can
It realizes by means of software and necessary general hardware platform, naturally it is also possible to pass through hardware.Based on this understanding, on
Stating technical solution, substantially the part that contributes to existing technology can be embodied in the form of software products in other words, should
Computer software product may be stored in a computer readable storage medium, such as ROM/RAM, magnetic disk, CD, including several fingers
It enables and using so that a computer equipment (can be personal computer, server or the network equipment etc.) executes each implementation
Method described in certain parts of example or embodiment.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although
Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used
To modify the technical solutions described in the foregoing embodiments or equivalent replacement of some of the technical features；
And these are modified or replaceed, technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution spirit and
Range.
Claims (10)
1. a kind of cell image segmentation method characterized by comprising
Obtain cell image to be split；
The cell image to be split is input in cell segmentation model, the cell point of the cell segmentation model output is obtained
Cut result；Wherein, the cell segmentation model is based on sample cell image and the corresponding true mark of the sample cell image
Image is signed, enhancing U-Net network is trained；The enhancing U-Net network is increased in initial U-Net network
The network number of plies, and BN layers of neural network is added in a network.
2. cell image segmentation method according to claim 1, which is characterized in that described by the cell image to be split
It is input in cell segmentation model, obtains the cell segmentation of the cell segmentation model output as a result, before further include:
Based on the sample cell image and the corresponding true tag image of the sample cell image, to the enhancing U-Net
Network is trained.
3. cell image segmentation method according to claim 2, which is characterized in that described to be based on the sample cell image
True tag image corresponding with the sample cell image is trained the enhancing U-Net network, before further include:
Any data set cell image and the corresponding data set label image of any data set cell image are merged,
It obtains data set and merges image；
Image is merged to the data set and carries out data enhancing, obtains data set enhancing image；The data enhancing includes geometry
At least one of transformation, Stochastic Elasticity deformation, random scaling, shear transformation counterclockwise；
Data set enhancing image is subjected to image separation, obtains the sample cell image and the sample cell image pair
The true tag image answered.
4. cell image segmentation method according to claim 2, which is characterized in that described to be based on the sample cell image
True tag image corresponding with the sample cell image is trained the enhancing U-Net network, before further include:
The sample cell image is normalized.
5. cell image segmentation method according to claim 2, which is characterized in that described to be based on the sample cell image
True tag image corresponding with the sample cell image is trained the enhancing U-Net network, before further include:
The enhancing U-Net network is initialized based on Gaussian Profile initial method.
6. cell image segmentation method according to claim 2, which is characterized in that described to be based on the sample cell image
True tag image corresponding with the sample cell image is trained the enhancing U-Net network, specifically includes:
The sample cell image and the corresponding true tag image of the sample cell image are input to the enhancing U-Net
Network obtains the training cell segmentation result of the enhancing U-Net network output；
The true tag image and the trained cell segmentation result are input to logarithm loss function, obtain loss function
Value；
Based on the loss function value, parameter regulation is carried out to the enhancing U-Net network by Adam optimization algorithm.
7. cell image segmentation method according to any one of claims 1 to 6, which is characterized in that the general is described wait divide
It cuts cell image to be input in cell segmentation model, obtains the cell segmentation of the cell segmentation model output as a result, before also
Include:
Test cell image is input in the cell segmentation model, the test cell of the cell segmentation model output is obtained
Segmentation result；
If the corresponding benchmark cell segmentation result of the test cell image is different from the test cell segmentation result, it is based on
The test cell image and the benchmark cell segmentation result training cell segmentation model.
8. a kind of Methods of Segmentation On Cell Images device characterized by comprising
Image acquisition unit, for obtaining cell image to be split；
Image segmentation unit obtains the cell point for the cell image to be split to be input in cell segmentation model
Cut the cell segmentation result of model output；Wherein, the cell segmentation model is thin based on sample cell image and the sample
The corresponding true tag image of born of the same parents' image is trained enhancing U-Net network；The enhancing U-Net network be
Increase the network number of plies in initial U-Net network, and BN layers of neural network is added in a network.
9. a kind of electronic equipment including memory, processor and stores the calculating that can be run on a memory and on a processor
Machine program, which is characterized in that the processor realizes cell as described in any one of claim 1 to 7 when executing described program
The step of image partition method.
10. a kind of non-transient computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer
The step of cell image segmentation method as described in any one of claim 1 to 7 is realized when program is executed by processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910504393.2A CN110288605A (en) | 2019-06-12 | 2019-06-12 | Cell image segmentation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910504393.2A CN110288605A (en) | 2019-06-12 | 2019-06-12 | Cell image segmentation method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110288605A true CN110288605A (en) | 2019-09-27 |
Family
ID=68003783
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910504393.2A Pending CN110288605A (en) | 2019-06-12 | 2019-06-12 | Cell image segmentation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110288605A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112102323A (en) * | 2020-09-17 | 2020-12-18 | 陕西师范大学 | Adherent nucleus segmentation method based on generation of countermeasure network and Caps-Unet network |
| CN112446427A (en) * | 2020-11-27 | 2021-03-05 | 王伟佳 | Method and device for identifying myeloid blood cells, storage medium and electronic equipment |
| CN112561897A (en) * | 2020-12-22 | 2021-03-26 | 电子科技大学中山学院 | Photonic crystal fiber end face structure extraction method based on U-Net |
| CN113128455A (en) * | 2021-04-30 | 2021-07-16 | 上海睿钰生物科技有限公司 | Cell image reconstruction model training method and system |
| CN113139973A (en) * | 2021-04-01 | 2021-07-20 | 武汉市疾病预防控制中心 | Artificial intelligence-based plasmodium identification method and equipment |
| CN114612738A (en) * | 2022-02-16 | 2022-06-10 | 中国科学院生物物理研究所 | Training method of cell electron microscope image segmentation model and organelle interaction analysis method |
| US11410303B2 (en) | 2019-04-11 | 2022-08-09 | Agilent Technologies Inc. | Deep learning based instance segmentation via multiple regression layers |
| US11803963B2 (en) | 2019-02-01 | 2023-10-31 | Sartorius Bioanalytical Instruments, Inc. | Computational model for analyzing images of a biological specimen |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180108139A1 (en) * | 2016-10-19 | 2018-04-19 | U.S. Department Of Veterans Affairs | System And Method For N-Dimensional Image Segmentation Using Convolutional Neural Networks |
| CN109064443A (en) * | 2018-06-22 | 2018-12-21 | 哈尔滨工业大学 | A kind of multi-model organ segmentation method and system based on abdominal ultrasound images |
| CN109191472A (en) * | 2018-08-28 | 2019-01-11 | 杭州电子科技大学 | Based on the thymocyte image partition method for improving U-Net network |
| 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 |
-
2019
- 2019-06-12 CN CN201910504393.2A patent/CN110288605A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180108139A1 (en) * | 2016-10-19 | 2018-04-19 | U.S. Department Of Veterans Affairs | System And Method For N-Dimensional Image Segmentation Using Convolutional Neural Networks |
| CN109064443A (en) * | 2018-06-22 | 2018-12-21 | 哈尔滨工业大学 | A kind of multi-model organ segmentation method and system based on abdominal ultrasound images |
| CN109191472A (en) * | 2018-08-28 | 2019-01-11 | 杭州电子科技大学 | Based on the thymocyte image partition method for improving U-Net network |
| 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 |
Non-Patent Citations (1)
| Title |
|---|
| 贝琛圆 等: "基于改进U-Net网络的腺体细胞图像分割算法", 《电子科技》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11803963B2 (en) | 2019-02-01 | 2023-10-31 | Sartorius Bioanalytical Instruments, Inc. | Computational model for analyzing images of a biological specimen |
| US11410303B2 (en) | 2019-04-11 | 2022-08-09 | Agilent Technologies Inc. | Deep learning based instance segmentation via multiple regression layers |
| US11748881B2 (en) | 2019-04-11 | 2023-09-05 | Agilent Technologies, Inc. | Deep learning based instance segmentation via multiple regression layers |
| CN112102323A (en) * | 2020-09-17 | 2020-12-18 | 陕西师范大学 | Adherent nucleus segmentation method based on generation of countermeasure network and Caps-Unet network |
| CN112102323B (en) * | 2020-09-17 | 2023-07-07 | 陕西师范大学 | Adhesion cell nucleus segmentation method based on generation of countermeasure network and Caps-Unet network |
| CN112446427A (en) * | 2020-11-27 | 2021-03-05 | 王伟佳 | Method and device for identifying myeloid blood cells, storage medium and electronic equipment |
| CN112446427B (en) * | 2020-11-27 | 2021-07-27 | 王伟佳 | Method and device for identifying myeloid blood cells, storage medium and electronic equipment |
| CN112561897A (en) * | 2020-12-22 | 2021-03-26 | 电子科技大学中山学院 | Photonic crystal fiber end face structure extraction method based on U-Net |
| CN113139973A (en) * | 2021-04-01 | 2021-07-20 | 武汉市疾病预防控制中心 | Artificial intelligence-based plasmodium identification method and equipment |
| CN113128455A (en) * | 2021-04-30 | 2021-07-16 | 上海睿钰生物科技有限公司 | Cell image reconstruction model training method and system |
| CN114612738A (en) * | 2022-02-16 | 2022-06-10 | 中国科学院生物物理研究所 | Training method of cell electron microscope image segmentation model and organelle interaction analysis method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110288605A (en) | Cell image segmentation method and device | |
| CN109740686A (en) | A kind of deep learning image multiple labeling classification method based on pool area and Fusion Features | |
| CN110705403A (en) | Cell sorting method, cell sorting device, cell sorting medium, and electronic apparatus | |
| CN112668579A (en) | Weak supervision semantic segmentation method based on self-adaptive affinity and class distribution | |
| US20220343683A1 (en) | Expression Recognition Method and Apparatus, Computer Device, and Readable Storage Medium | |
| CN109117781A (en) | Method for building up, device and the more attribute recognition approaches of more attribute Recognition Models | |
| CN111539480A (en) | Multi-class medical image identification method and equipment | |
| CN111882548A (en) | Method and device for counting cells in pathological image based on deep learning | |
| CN110427484A (en) | A kind of Chinese natural language processing method based on deep learning | |
| CN113378940B (en) | Neural network training method and device, computer equipment and storage medium | |
| CN113128478A (en) | Model training method, pedestrian analysis method, device, equipment and storage medium | |
| CN113706151A (en) | Data processing method and device, computer equipment and storage medium | |
| CN112052687B (en) | Semantic feature processing method, device and medium based on depth separable convolution | |
| CN104751175A (en) | Multi-label scene classification method of SAR (Synthetic Aperture Radar) image based on incremental support vector machine | |
| CN111178196B (en) | Cell classification method, device and equipment | |
| CN113010683A (en) | Entity relationship identification method and system based on improved graph attention network | |
| Khashman | Investigation of different neural models for blood cell type identification | |
| CN115035334A (en) | Multi-classification change detection method and system for multi-scale fusion double-temporal remote sensing image | |
| CN110807159B (en) | Data marking method and device, storage medium and electronic equipment | |
| CN108805152A (en) | A kind of scene classification method and device | |
| CN110348023A (en) | A kind of method, apparatus, storage medium and the electronic equipment of Chinese text participle | |
| CN110688880A (en) | License plate identification method based on simplified ResNet residual error network | |
| CN113297376A (en) | Legal case risk point identification method and system based on meta-learning | |
| CN113627522A (en) | Image classification method, device and equipment based on relational network and storage medium | |
| CN112860843A (en) | News long text sentiment analysis method and device |
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: 20190927 |