CN109872307A - Method, relevant device and the medium of lump in a kind of detection biological tissue images - Google Patents
Method, relevant device and the medium of lump in a kind of detection biological tissue images Download PDFInfo
- Publication number
- CN109872307A CN109872307A CN201910092954.2A CN201910092954A CN109872307A CN 109872307 A CN109872307 A CN 109872307A CN 201910092954 A CN201910092954 A CN 201910092954A CN 109872307 A CN109872307 A CN 109872307A
- Authority
- CN
- China
- Prior art keywords
- parting
- biological tissue
- image
- target
- lump
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- 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
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/155—Segmentation; Edge detection involving morphological operators
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20028—Bilateral filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30068—Mammography; Breast
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30096—Tumor; Lesion
Abstract
This application discloses a kind of methods of lump in detection biological tissue images, comprising: obtains target biological tissue image；Determine the parting type of target biological tissue image；Determine that target object and target object in target biological tissue image are the probability of lump；The probability that target object is lump is compared with probability threshold value corresponding to the parting type of target biological tissue image, probability threshold value corresponding to the biological tissue images of different parting types is different；If the probability that target object is lump is greater than probability threshold value corresponding to the parting type of target biological tissue image, detect that target object is lump.Technical scheme is due to when detecting biological tissue images, such as: when detection galactophore image, it can detecte out the parting type of biological tissue images, there is different lump probability threshold values in the biological tissue of different parting types, to improve the accuracy of Mass detection.
Description
Technical field
This application involves field of computer technology, and in particular to a kind of to detect the method for lump in biological tissue images, instruction
Practice method, related device, equipment and the medium of model.
Background technique
Breast molybdenum target (mammograms) is widely used in breast cancer early screening.Lump is to judge whether mammary gland is normal
Important local feature wherein the positioning of doubtful Malignant mass can provide preferable good pernicious judgment basis for doctor be cream
One of the most important clue of gland cancer diagnosis.
The form of lump and not of uniform size, and usually show lower contrast.Traditional lump Detection Techniques are main
Supervision detection is carried out according to the priori features of lump or carries out non-supervisory point using the severability of lump and its hetero-organization
It cuts, it is often necessary to the problem of facing lump location difficulty.
The prior art is partitioned into suspicious region using partitioning algorithm, then carries out lump screening according to suspicious region form,
The accuracy of screening is very low.
Summary of the invention
The embodiment of the present application provides a kind of method for detecting lump in biological tissue images, and biological tissue images can be improved
The accuracy of middle Mass detection.The embodiment of the present application also provides corresponding device and storage mediums.
The application first aspect provides a kind of method for detecting lump in biological tissue images, comprising:
Obtain target biological tissue image；
Determine the parting type of the target biological tissue image；
Determine that target object and the target object in the target biological tissue image are the probability of lump；
It will be general corresponding to parting type of the probability with the target biological tissue image that the target object is lump
Rate threshold value is compared, and probability threshold value corresponding to the biological tissue images of different parting types is different；
If the probability that the target object is lump is greater than corresponding to the parting type of the target biological tissue image
Probability threshold value then detects that the target object is lump.
In a kind of possible implementation, the biological tissue images are galactophore image, the determination target organism
The parting type of organization chart picture may include:
The parting type of target galactophore image is determined by target mammary gland parting model, the target mammary gland parting model is
It is trained by multiple galactophore images of different parting types and the parting type information of each galactophore image.
In a kind of possible implementation, the method for lump can also include: in the detection galactophore image
Mark the lump being detected；
Output is comprising with markd galactophore image.
In a kind of possible implementation, the difference parting types include at least two in following parting type: rouge
Fat type, few body of gland type, polyadenous figure and dense form；
The corresponding probability threshold value of the lard type is a, few corresponding probability threshold value of body of gland type is b, the polyadenous body
The corresponding probability threshold value of type is c, and the corresponding probability threshold value of the dense form is d, and described a, b, c and d are both greater than 0, and a <b < c <
d。
In a kind of possible implementation, the target object in the determination target biological tissue image, Yi Jisuo
The probability that target object is lump is stated, may include:
The target galactophore image is pre-processed and divided, divides subgraph to determine；
The segmentation subgraph is inputted into disaggregated model, determines target sub-object included in each segmentation subgraph,
The target sub-object is contained in the target object；
Determine that the target object is the probability of lump according to each target sub-object.
It is described to determine that the target object is the general of lump according to each target sub-object in a kind of possible implementation
Rate may include:
Merge region overlapped in each target sub-object；
According to the set of the target sub-object after merging, determine that the target object is the probability of lump.
The application second aspect provides a kind of method of training pattern, comprising:
Sample image set is obtained, the sample image set includes multiple biological tissue images of different parting types,
And the parting type information of each biological tissue images；
Initial biological tissue's parting model is trained by described image set, with the determination initial biological tissue
The reference parameter of parting model；
The reference parameter is inputted into initial biological tissue's parting model, is used for target biological tissue parting to determine
Model, the target biological tissue parting model are used to determine the parting type of biological tissue images.
In a kind of possible implementation, the biological tissue images are galactophore image, described to pass through described image set
It is defeated that initial mammary gland parting model is trained, with the reference parameter of the determination initial mammary gland parting model, may include:
Extract the characteristic information of each galactophore image；
By the characteristic information of each galactophore image, and parting type information corresponding with the galactophore image is as one
Group training parameter；
By multiple galactophore images, respectively corresponding one group of training parameter instructs the initial mammary gland parting model
Practice, with the reference parameter of the determination initial mammary gland parting model.
The application third aspect provides a kind of device for detecting lump in biological tissue images, comprising:
Acquiring unit, for obtaining target biological tissue image；
First determination unit, the parting type of the target biological tissue image for determining the acquiring unit acquisition；
Second determination unit, the target object in target biological tissue image for determining the acquiring unit acquisition,
And the target object is the probability of lump；
Comparing unit, for by second determination unit determine target object be lump probability with described first really
Probability threshold value corresponding to the parting type for the target biological tissue image that order member determines is compared, different parting types
Probability threshold value corresponding to biological tissue images is different；
Detection unit, if comparing the probability that target object is lump for the comparing unit is greater than the target organism
When probability threshold value corresponding to the parting type of organization chart picture, then detect that the target object is lump.
In a kind of possible implementation, first determination unit, for being mammary gland figure in the biological tissue images
When picture, the parting type of target galactophore image is determined by target mammary gland parting model, the target mammary gland parting model is logical
What the parting type information of the multiple galactophore images and each galactophore image of crossing different parting types was trained.
In a kind of possible implementation, described device can also include:
Marking unit, for marking the lump being detected；
Output unit includes the galactophore image with marking unit label for exporting.
In a kind of possible implementation, the difference parting types include at least two in following parting type: rouge
Fat type, few body of gland type, polyadenous figure and dense form；
The corresponding probability threshold value of the lard type is a, few corresponding probability threshold value of body of gland type is b, the polyadenous body
The corresponding probability threshold value of type is c, and the corresponding probability threshold value of the dense form is d, and described a, b, c and d are both greater than 0, and a <b < c <
d。
In a kind of possible implementation, second determination unit is used for:
The target galactophore image is pre-processed and divided, divides subgraph to determine；
The segmentation subgraph is inputted into disaggregated model, determines target sub-object included in each segmentation subgraph,
The target sub-object is contained in the target object；
Determine that the target object is the probability of lump according to each target sub-object.
In a kind of possible implementation, second determination unit is used for:
Merge region overlapped in each target sub-object；
According to the set of the target sub-object after merging, determine that the target object is the probability of lump.
The application fourth aspect provides a kind of device of training pattern, comprising:
Acquiring unit, for obtaining sample image set, the sample image set includes the multiple of different parting types
The parting type information of biological tissue images and each biological tissue images；
Training unit, for being trained by described image set to initial biological tissue's parting model, to determine
State the reference parameter of initial biological tissue's parting model；
Determination unit is used for mesh for the reference parameter to be inputted initial biological tissue's parting model to determine
Biological tissue's parting model is marked, the target biological tissue parting model is used to determine the parting type of biological tissue images.
In a kind of possible implementation, the training unit is used for: when the biological tissue images are galactophore image,
Extract the characteristic information of each galactophore image；
By the characteristic information of each galactophore image, and parting type information corresponding with the galactophore image is as one
Group training parameter；
By multiple galactophore images, respectively corresponding one group of training parameter instructs the initial mammary gland parting model
Practice, with the reference parameter of the determination initial mammary gland parting model.
The 5th aspect of the embodiment of the present application provides a kind of computer equipment, and the computer equipment includes: input/output
(I/O) interface, processor and memory are stored with program instruction in the memory；
The processor executes such as above-mentioned first aspect or first aspect for executing the program instruction stored in memory
Method described in any one possible implementation.
The 6th aspect of the embodiment of the present application provides a kind of computer equipment, and the computer equipment includes: input/output
(I/O) interface, processor and memory are stored with program instruction in the memory；
The processor executes such as above-mentioned second aspect or second aspect for executing the program instruction stored in memory
Method described in any one possible implementation.
The 7th aspect of the embodiment of the present application provides a kind of computer readable storage medium, including instruction, when described instruction exists
When running in computer equipment, so that the computer equipment executes such as above-mentioned first aspect or first aspect, any one may
Implementation described in method.
The embodiment of the present application eighth aspect provides a kind of computer readable storage medium, including instruction, when described instruction exists
When running in computer equipment, so that the computer equipment executes such as above-mentioned second aspect or second aspect, any one may
Implementation described in method.
The 9th aspect of the application provides a kind of computer program product comprising instruction, when it runs on computers
When, so that computer executes method described in above-mentioned first aspect or first aspect any one possible implementation.
The tenth aspect of the application provides a kind of computer program product comprising instruction, when it runs on computers
When, so that computer executes method described in above-mentioned second aspect or second aspect any one possible implementation.
The tenth one side of the application provides a kind of medical image detection system, and the medical image detection system includes image
Scanning device and image processing equipment；
Image scanning apparatus sends the medical image for scanning medical image, and to described image processing equipment；
Image processing equipment is for executing method described in any one of first aspect, or execution as appointed in second aspect
Method described in one.
Scheme provided by the embodiments of the present application can detecte out biological tissue images when detecting biological tissue images
Parting type, there is different lump probability threshold values in the biological tissue of different parting types, to improve the accurate of Mass detection
Degree.
Detailed description of the invention
Fig. 1 is a Sample Scenario schematic diagram of training objective mammary gland parting model in the embodiment of the present application；
Fig. 2 is an embodiment schematic diagram of the method for training pattern in the embodiment of the present application；
Fig. 3 is the schematic diagram of a scenario that lump in galactophore image is detected in the embodiment of the present application；
Fig. 4 is another schematic diagram of a scenario that lump in galactophore image is detected in the embodiment of the present application；
Fig. 5 is the embodiment of the method schematic diagram that lump in galactophore image is detected in the embodiment of the present application；
Fig. 6 is the another method embodiment schematic diagram that lump in galactophore image is detected in the embodiment of the present application；
Fig. 7 is an embodiment schematic diagram of the device of training pattern in the embodiment of the present application；
Fig. 8 is the embodiment schematic diagram that the device of lump in galactophore image is detected in the embodiment of the present application；
Fig. 9 is another embodiment schematic diagram that the device of lump in galactophore image is detected in the embodiment of the present application；
Figure 10 is an embodiment schematic diagram of computer equipment in the embodiment of the present application.
Specific embodiment
With reference to the accompanying drawing, embodiments herein is described, it is clear that described embodiment is only the application
The embodiment of a part, instead of all the embodiments.Those of ordinary skill in the art are it is found that with the development of technology and new field
The appearance of scape, technical solution provided by the embodiments of the present application are equally applicable for similar technical problem.
The embodiment of the present application provides a kind of method for detecting lump in biological tissue images, and biological tissue images can be improved
The accuracy of middle Mass detection.The embodiment of the present application also provides corresponding device and storage mediums.It carries out individually below detailed
Explanation.
With the development of artificial intelligence technology, machine can be with auxiliary judgment biological tissue images, such as in galactophore image are
The no technology for having lump, how much differences of the embodiment of the present application in view of body of gland on everyone mammary gland, the supramammary gland of somebody
Body is more, and the supramammary body of gland of somebody is few, and when judging supramammary lump, it will receive the how many influence of body of gland.So this
Apply in embodiment, when determining the lump in galactophore image, it is contemplated that the parting type of galactophore image, in conjunction with galactophore image
Parting type go to determine the lump situation in the galactophore image again, the accuracy of Mass detection can be improved.About mammary gland figure
The determination of the parting type of picture can first train a mammary gland parting model, then to the mammary gland parting mode input mammary gland figure
Picture, so that it may export the parting type of the galactophore image.Mammary gland parting model can be a deep neural network model.About
Mammary gland parting model can be trained to obtain by the sample of the galactophore image of a large amount of different parting types.Certainly, this Shen
It please be only illustrated by taking galactophore image as an example in embodiment, in fact, can be by the biological tissue of parting type about other
Image belongs to the claimed scheme of the embodiment of the present application.
Below with reference to a schematic diagram of a scenario of the training pattern of Fig. 1, introduce in the embodiment of the present application about training mammary gland point
The process of pattern type.
As shown in Figure 1, may include: that computer is set in a scene embodiment of training pattern provided by the embodiments of the present application
Standby 10 obtain sample image set from database 20, which includes multiple mammary gland figures of different parting types
The parting type information of picture and each galactophore image.Data in sample image set can be collects in advance, each cream
The parting type information of gland image can be related fields expert mark.Galactophore image can be divided in the embodiment of the present application
For following several parting types:: lard type, few body of gland type, polyadenous figure and dense form.Certainly, it should be noted that the application
Embodiment provides a kind of thought of galactophore image parting Type division, the side of other related galactophore image parting Type divisions
Case, though divide type is different from the application or the title of parting type and the application difference, but as long as being related to cream
The division of gland image parting type, belongs to the claimed range of the embodiment of the present application.
Initial mammary gland parting model is configured in computer equipment 10, computer equipment 10 can pass through described image set
Initial mammary gland parting model is trained, with the reference parameter of the determination initial mammary gland parting model；By described with reference to ginseng
Number inputs the initial mammary gland parting model, is used for target mammary gland parting model to determine, the target mammary gland parting model is used
In the parting type for determining galactophore image.
In a kind of possible implementation, initial mammary gland parting model can be understood as the letter of a deep neural network
It counts, coefficient is in unknown state in the function, these coefficients for being in unknown state can be understood as initial mammary gland parting mould
The reference parameter of type, the characteristic information of each galactophore image can be understood as multiple input parameters, corresponding parting type letter
Breath can be understood as corresponding output parameter, and the relationship for inputting parameter and output parameter can be expressed as y=f (x1, x2 ...
xN).Many group input and output parameters are input in initial mammary gland parting model, so that it may determine that these are unknown
Reference parameter obtains target mammary gland parting model to complete the training of model.In this way, defeated into target mammary gland parting model
After entering a galactophore image, so that it may by extracting the characteristic information in the galactophore image (x1, x2 ... xN), must export pair
The output parameter y answered realizes the parting type detection to galactophore image.
Correspondingly, in conjunction with above-mentioned Sample Scenario shown in FIG. 1, as shown in Fig. 2, training pattern provided by the embodiments of the present application
An embodiment of method may include:
101, sample image set is obtained, the sample image set includes multiple biological tissues figure of different parting types
The parting type information of picture and each biological tissue images.
102, initial biological tissue's parting model is trained by described image set, with the determination just eozoon
The reference parameter of tissue typing model.
In a kind of possible implementation, which may include: if biological tissue images are galactophore image
Extract the characteristic information of each galactophore image；
By the characteristic information of each galactophore image, and parting type information corresponding with the galactophore image is as one
Group training parameter；
By the multiple galactophore image respectively corresponding one group of training parameter to the initial mammary gland parting model into
Row training, with the reference parameter of the determination initial mammary gland parting model.
It certainly, can also be using above-mentioned process training reference parameter if biological tissue images are the image at other positions.
103, the reference parameter is inputted into initial biological tissue's parting model, is used for target biological tissue to determine
Parting model, the target biological tissue parting model are used to determine the parting type of biological tissue images.
The method of training pattern provided by the embodiments of the present application can train the parting for determining biological tissue images
The target biological tissue parting model of type, the training process about target biological tissue parting model can participate in above-mentioned Fig. 1
Partial description is understood that it is no longer repeated herein.
From the description above it is found that target biological tissue parting model can be trained, certainly, mesh can be also trained
Mammary gland parting model training is marked, then the target mammary gland parting model can be used first true when lump in detecting galactophore image
The parting type for making target galactophore image detected, then detects lump again.It is usually to pass through analysis to swell when Mass detection
The feature in block region determines that the region is the probability of lump, if the lump in the galactophore image of either which kind of parting type is sentenced
Disconnected is all that then undoubtedly will cause accuracy of identification decline using the comparison threshold value passed through.Therefore, every kind is directed in the embodiment of the present application
The galactophore image of parting type is all configured with different size of for judging the probability threshold value of lump.
Such as: galactophore image, which is described above, in the embodiment of the present application can be divided into lard type, few body of gland type, polyadenous
Figure and dense form.Then the corresponding probability threshold value of the lard type can be with for a, few corresponding probability threshold value of body of gland type
For b, the corresponding probability threshold value of the polyadenous figure can be c, and the corresponding probability threshold value of the dense form can be d, a,
B, c and d are both greater than 0.Because of lard type, body of gland is seldom, is easier to detect the lump in galactophore image, so a is not needed
What is be arranged is too big, and when lump in these types of parting type in the galactophore image of lard type is easiest to detection, so a is most
It is small, and so on, then there is a <b < c < d.Value about a, b, c and d can be through empirical value setting, be also possible to pass through
What some algorithms were calculated, the method for determination of the value of a, b, c and d is not limited in the embodiment of the present application.
In this way, when detecting the lump in galactophore image, so that it may be respectively adopted corresponding with the parting type of galactophore image
Probability threshold value go to make a decision, so as to improve the accuracy of Mass detection.It is introduced in the embodiment of the present application below with reference to Fig. 3
Detect the process of lump in galactophore image.
Fig. 3 is the schematic diagram of a scenario that lump in galactophore image is detected in the embodiment of the present application.
As shown in figure 3, may include to detect in the embodiment of the present application in galactophore image in a scene embodiment of lump
Image capture device 30, computer equipment 40 and display 50, image capture device 30, computer equipment 40 and display 50 it
Between image transmitting can be carried out by network.Image capture device 30 can be molybdenum target machine, which can lead to
The target galactophore image of over-discharge radiography acquisition patient, and by the target galactophore image by network transmission to computer equipment 40,
Computer equipment 40 can determine the parting type of the target galactophore image；Then the mesh in the target galactophore image is determined again
It marks object and the target object is the probability of lump；It is the probability and the target mammary gland of lump by the target object
Probability threshold value corresponding to the parting type of image is compared, probability threshold value corresponding to the galactophore image of different parting types
It is different；If the probability that the target object is lump is greater than probability threshold corresponding to the parting type of the target galactophore image
Value then detects that the target object is lump.
In a kind of possible implementation, configured with the target in embodiment corresponding to above-mentioned Fig. 1 in computer equipment 40
Mammary gland parting model can determine the parting type of target galactophore image by target mammary gland parting model.
After computer equipment 40 detects lump, the lump being detected can be marked, and export comprising with label
Galactophore image.In Fig. 3 on display 50 as shown in galactophore image in if marked the position of lump.Certainly, lump
Mark mode can there are many kinds of, this being not limited in Fig. 3 is a kind of.
Certainly, the scheme of lump is also not necessarily limited to only use Fig. 3 institute in detection galactophore image provided by the embodiment of the present application
The collection in worksite galactophore image shown, field assay go out the scene of result, if piece of only one breast cancer, can also pass through
To the Image Acquisition of piece, scheme provided by Lai Shixian the embodiment of the present application.
As shown in figure 4, image is adopted to detect in galactophore image in another scene embodiment of lump in the embodiment of the present application
Integrating equipment 30 can be filming instrument, can acquire the image in piece of breast cancer and be transmitted further to computer equipment 40, meter
Subsequent process and the process described in embodiment corresponding to above-mentioned Fig. 3 for calculating the execution of equipment 40 are essentially identical, no longer heavy herein
It repeats again.
Certainly, scene corresponding to above-mentioned Fig. 3 and Fig. 4 is two kinds of examples, if computer equipment and aobvious in other scenes
Show that device integrates, as long as having the function of computer equipment 40 in scene corresponding to above-mentioned Fig. 3 and Fig. 4, can also pass through
One equipment realizes the display of lump and output result in above-mentioned detection galactophore image.If image collecting function also with computer
Equipment and display are integrated on one device, then only needing to realize that above-mentioned Fig. 3 and Fig. 4 institute is right by an equipment
Answer the process of acquisition, detection and the display of the image in scene.
Above-mentioned Fig. 3 and scene shown in Fig. 4 are illustrated by taking breast cancer as an example, in fact, the application is implemented
Example is not limited to breast cancer,
As shown in figure 5, an embodiment of the method for lump can in detection biological tissue images provided by the embodiments of the present application
To include:
201, target biological tissue image is obtained.
Biological tissue images for example can be target galactophore image, and the acquisition modes of target galactophore image can participate in above-mentioned
Acquisition modes in two kinds of scenes of Fig. 3 or Fig. 4 are understood.
202, the parting type of the target biological tissue image is determined.
Such as: the parting type of target galactophore image can be with are as follows: lard type, few body of gland type, polyadenous figure or dense form.
203, determine that target object and the target object in the target biological tissue image are the general of lump
Rate.
Lump in the embodiment of the present application can be doubtful Malignant mass, certainly, be also not necessarily limited to doubtful Malignant mass, can also
It can be benign tumors.
In a kind of possible implementation, when biological tissue images are galactophore image, which may include:
The target galactophore image is pre-processed and divided, divides subgraph to determine；
The segmentation subgraph is inputted into disaggregated model, determines target sub-object included in each segmentation subgraph,
The target sub-object is contained in the target object；
Determine that the target object is the probability of lump according to each target sub-object.
Wherein, described to determine that the target object is the probability of lump according to each target sub-object, may include:
Merge region overlapped in each target sub-object；
According to the set of the target sub-object after merging, determine that the target object is the probability of lump.
The merging gone out is actually to be directed to overlapped two parts, to remove a portion.
It 204, will be corresponding to the parting type of probability and the target biological tissue image that the target object is lump
Probability threshold value be compared, probability threshold value corresponding to the biological tissue images of different parting types is different.
The step 204 can participate in the following table 1 and be understood by taking galactophore image as an example:
Table 1
If the parting type of target galactophore image is lard type, E is compared with a, if point of target galactophore image
Type type is few body of gland type, then is compared E with b, if the parting type of target galactophore image is polyadenous figure, by E and c
It is compared, if the parting type of target galactophore image is dense form, E is compared with d.
If 205, the target object is right for the parting type that the probability of lump is greater than the target biological tissue image
The probability threshold value answered then detects that the target object is lump.
If E=0.45, a=0.4, b=0.5, c=0.6, d=0.7, then the parting type of target galactophore image is fat
When type, E > a, it is determined that the target object is lump.If the parting type of target galactophore image is few body of gland type, polyadenous figure
Or dense form, E are respectively smaller than b, c and d, it is determined that going out the target object is not lump.
In the embodiment of the present application, when detecting galactophore image, it can detecte out the parting type of galactophore image, different partings
The mammary gland of type has different lump probability threshold values, to improve the accuracy of Mass detection.
It can also include: the lump that label is detected after step 205 in a kind of possible implementation；Output packet
Containing with markd galactophore image.The corresponding contents that this possible implementation can participate in Fig. 3 or Fig. 4 are understood,
Repetition is no longer done herein to repeat.
Below with reference to Fig. 6, by taking galactophore image as an example, lump in detection galactophore image provided by the embodiments of the present application is introduced
Another embodiment of method.
As shown in fig. 6, another embodiment of the method for lump can be in detection galactophore image provided by the embodiments of the present application
Include:
In a branch, molybdenum target picture, that is, target galactophore image are input to mammary gland parting network, the mammary gland point
The target mammary gland parting model of type network namely above.Mammary gland parting network is to be trained to obtain by training sample
, the target mammary gland parting model of the mammary gland parting network namely above, specific training process can participate in corresponding to Fig. 1
Embodiment understood.
In another branch, molybdenum target picture, that is, target galactophore image are pre-processed, to pretreated figure
As being split.
Liang Ge branch combines, and the process of lump judgement can be in conjunction with point for determining target galactophore image by mammary gland parting network
Probability threshold value corresponding to type type obtains testing result after lump judgement.
Wherein, pretreated process may include:
(1), it normalizes
Image grayscale range is stretched to 0-255 by linear stretch, improves the robustness of subsequent processing.
(2), breast area is divided
Operation is opened using morphology and binaryzation extracts breast area, and the backgrounds such as removal label, opening operation can remove carefully
Broken tissue and noise, cutting procedure carry out two classification using big saliva segmentation (Otsu) method, can effectively extract breast tissue area
Domain.
(3), histogram equalization
Subsequent singulation algorithm is carried out based on image histogram, it is therefore desirable to improve subsequent processing robust by histogram equalization
Property.
(4), bilateral filtering
Using noise that may be present in bilateral filtering removal breast tissue, and region homogeneity is improved to a certain extent
Property, in addition bilateral filtering will not destroy segmenting edge.
Image segmentation may include:
(1), gene genetic is divided
Dimension is reduced using two-dimensional wavelet transformation (series 3) to target galactophore image, for low detail pictures, in normalizing
Its image histogram is counted after change, and carries out image segmentation according to this histogram.Gene genetic is used for the segmentation of histogram
Algorithm, gene use binary coded form, and length is equal to gray level number, represents the gray level when place value is 0 as segmentation
Threshold value.Genetic algorithm cost function is calculated using maximum between-cluster variance and minimum variance within clusters as standard using general gene genetic
Method process, after initialization of population iteration select, intersect, three processes that make a variation until convergence (initial population quantity for
30, the number of iterations 40, selection rate 10%, crossing-over rate 80%, aberration rate 10%), segmentation threshold is finally exported, according to
This threshold value is split operation to original image.
(2), morphology opens operation
Operation is opened using morphology to segmented image, thymus gland connection etc. is disconnected, subsequent sections is facilitated to extract.
(3), region unit extracts
For segmentation result, extraction gray level first is higher, such as: the region of Top5 gray level, for meeting condition
Region, biggish 10 regions of every molybdenum target image selection area are as candidate region.
Lump judges
(1), neural metwork training and classification
It using domestic hospitals data, engages expert's labeled data (2200+), doubtful Malignant mass is remaining as positive sample
Obvious benign tumors and background area as negative sample, (due to being molybdenum target picture, be substantially carried out and turn over after being enhanced by data
The data enhancing for turning and cutting, enhances, in addition inputting doubtful Malignant mass sample must wrap without the data for carrying out color space
Containing entire lump region and there is a small amount of background area to surround), as training data of the present invention input Google publication
InceptionV3 model is trained, and the output classification number of model is re-set as 2.Wherein the weights initialisation of model is first
Using ImageNet data set, public data collection DDSM is then used, finally carries out transfer learning using training data of the present invention
To the end Model Weight (descent algorithm use RMSprop, batch processing size be 64, initial learning rate be 0.01, maximum changes
100000) generation number is.After the completion of model training, for the candidate area blocks arbitrarily inputted, it can be obtained by network query function
Its whether be doubtful Malignant mass probability tag, usually, it 0.5 is considered as doubtful Malignant mass that probability, which is greater than,.
(2), non-maxima suppression
For being judged as the region of doubtful Malignant mass, using the region of non-maxima suppression method removal overlapping, wherein
Degree of overlapping threshold value is set as 50%, and main purpose is to reduce rate of false alarm, and the standard of doubtful Malignant mass positioning can be improved at the same time
True property.
Mammary gland parting may include:
(1), mammary gland parting network training and classification
Using domestic hospitals data, expert is engaged to be labelled with mammary gland typing data (6000+), the training stage uses Google
The InceptionV3 deep learning network of publication is trained, and the output classification of model is re-set as 4, and wherein model is first
Beginningization weight use ImageNet data set, descent algorithm use adam, batch processing size 64, initial learning rate 0.0001, most
Big the number of iterations 100.Application stage inputs mammary gland parting network, obtains mammary gland parting for the molybdenum target picture of input system
Classification.
(2), lump decision threshold selects
According to the mammary gland parting that classification obtains, corresponding lump is returned respectively and determines classification thresholds, corresponding to lard type, less
Body of gland type, polyadenous figure, dense form breast, threshold value are respectively a, b, c, d, wherein a <b < c < d.Breast i.e. more for body of gland,
It is determined as that the threshold value of lump correspondinglys increase, is conducive to the interference for avoiding lumps body of gland from positioning lump.
More than, it describes training pattern and detects the scheme of the lump in galactophore image, introduce this Shen with reference to the accompanying drawing
It please related device in embodiment.
As shown in fig. 7, an embodiment of the device 70 of training pattern provided by the embodiments of the present application may include:
Acquiring unit 701, for obtaining sample image set, the sample image set includes the more of different parting types
The parting type information of a biological tissue images and each biological tissue images；
Training unit 702, the image collection for being obtained by the acquiring unit 701 is to initial biological tissue's parting mould
Type is trained, with the reference parameter of determination initial biological tissue's parting model；
Determination unit 703, the reference parameter for training the training unit 702 input the initial biological tissue
Parting model is used for target biological tissue parting model to determine, the target biological tissue parting model is for determining biology
The parting type of organization chart picture.
The device of training pattern provided by the embodiments of the present application can train the parting for determining biological tissue images
The target biological tissue parting model of type improves biological tissue images so as to realize the lump judgement of parting type
The accuracy of middle lump identification.
In a kind of possible implementation, the training unit is used for: when biological tissue images are galactophore image,
Extract the characteristic information of each galactophore image；
By the characteristic information of each galactophore image, and parting type information corresponding with the galactophore image is as one
Group training parameter；
By the multiple galactophore image respectively corresponding one group of training parameter to the initial mammary gland parting model into
Row training, with the reference parameter of the determination initial mammary gland parting model.
As shown in figure 8, the embodiment provided by the embodiments of the present application for detecting the device 80 of lump in biological tissue images
May include:
Acquiring unit 801, for obtaining target biological tissue image；
First determination unit 802, the parting class of the target biological tissue image for determining the acquisition of acquiring unit 801
Type；
Second determination unit 803, the target in target biological tissue image for determining the acquisition of acquiring unit 801
Object and the target object are the probability of lump；
Comparing unit 804, for by second determination unit 803 determine target object be lump probability with it is described
Probability threshold value corresponding to the parting type for the target biological tissue image that first determination unit 802 determines is compared, different
Probability threshold value corresponding to the biological tissue images of parting type is different；
Detection unit 805, if comparing the probability that target object is lump for the comparing unit 804 is greater than the mesh
When marking probability threshold value corresponding to the parting type of biological tissue images, then detect that the target object is lump.
The device of lump in detection biological tissue images provided by the embodiments of the present application, when detecting biological tissue images,
Can detecte out the parting type of biological tissue images, there is different lump probability threshold values in the biological tissue of different parting types,
To improve the accuracy of Mass detection.
In a kind of possible implementation, first determination unit 802, for being galactophore image when biological tissue images
When, the parting type of target galactophore image is determined by target mammary gland parting model, the target mammary gland parting model is to pass through
What multiple galactophore images of different parting types and the parting type information of each galactophore image were trained.
As shown in figure 9, another embodiment provided by the embodiments of the present application for detecting the device 80 of lump in galactophore image is also
May include:
Marking unit 806, for marking the lump detected by the detection unit 805；
Output unit 807 includes the galactophore image with the marking unit 806 label for exporting.
In a kind of possible implementation, the corresponding probability threshold value of the lard type is a, few body of gland type is corresponding general
Rate threshold value is b, and the corresponding probability threshold value of the polyadenous figure is c, and the corresponding probability threshold value of the dense form is d, described a, b, c
0, and a <b < c < d are both greater than with d.
In a kind of possible implementation, second determination unit 803 is used for:
The target galactophore image is pre-processed and divided, divides subgraph to determine；
The segmentation subgraph is inputted into disaggregated model, determines target sub-object included in each segmentation subgraph,
The target sub-object is contained in the target object；
Determine that the target object is the probability of lump according to each target sub-object.
In a kind of possible implementation, second determination unit 803 is used for:
Merge region overlapped in each target sub-object；
According to the set of the target sub-object after merging, determine that the target object is the probability of lump.
Figure 10 is the structural schematic diagram of computer equipment 90 provided by the embodiments of the present application.The computer equipment 90 includes
Processor 910, memory 940 and input and output (I/O) interface 930, memory 940 may include read-only memory and deposit at random
Access to memory, and operational order and data are provided to processor 910.The a part of of memory 940 can also include non-volatile
Random access memory (NVRAM).
In some embodiments, memory 940 stores following element, executable modules or data structures, or
Their subset of person or their superset:
In the embodiment of the present application, during training pattern, processor 910 is by calling memory 940 to store
Operational order (operational order is storable in operating system) executes following process:
Sample image set is obtained, the sample image set includes multiple biological tissue images of different parting types,
And the parting type information of each biological tissue images；
Initial biological tissue's parting model is trained by described image set, with the determination initial biological tissue
The reference parameter of parting model；
The reference parameter is inputted into initial biological tissue's parting model, is used for target biological tissue parting to determine
Model, the target biological tissue parting model are used to determine the parting type of biological tissue images.
The scheme of training pattern provided by the embodiments of the present application can train the parting for determining biological tissue images
The target biological tissue parting model of type improves biological tissue's figure so as to realize the lump judgement of point parting type
The accuracy of lump identification as in.
Processor 910 controls the operation of computer equipment 90, and processor 910 can also be known as CPU (Central
Processing Unit, central processing unit).Memory 940 may include read-only memory and random access memory, and
Instruction and data is provided to processor 910.The a part of of memory 940 can also include nonvolatile RAM
(NVRAM).The various components of computer equipment 90 are coupled by bus system 920 in specific application, wherein bus
System 920 can also include power bus, control bus and status signal bus in addition etc. in addition to including data/address bus.But it is
For the sake of clear explanation, in figure various buses are all designated as bus system 920.
The method that above-mentioned the embodiment of the present application discloses can be applied in processor 910, or be realized by processor 910.
Processor 910 may be a kind of IC chip, the processing capacity with signal.During realization, the above method it is each
Step can be completed by the integrated logic circuit of the hardware in processor 910 or the instruction of software form.Above-mentioned processing
Device 910 can be general processor, digital signal processor (DSP), specific integrated circuit (ASIC), ready-made programmable gate array
(FPGA) either other programmable logic device, discrete gate or transistor logic, discrete hardware components.May be implemented or
Person executes disclosed each method, step and logic diagram in the embodiment of the present application.General processor can be microprocessor or
Person's processor is also possible to any conventional processor etc..The step of method in conjunction with disclosed in the embodiment of the present application, can be straight
Connect and be presented as that hardware decoding processor executes completion, or in decoding processor hardware and software module combination executed
At.Software module can be located at random access memory, and flash memory, read-only memory, programmable read only memory or electrically-erasable can
In the storage medium of this fields such as programmable memory, register maturation.The storage medium is located at memory 940, and processor 910 is read
Information in access to memory 940, in conjunction with the step of its hardware completion above method.
In a kind of possible implementation, processor 910 is used for: when biological tissue images are galactophore image,
Extract the characteristic information of each galactophore image；
By the characteristic information of each galactophore image, and parting type information corresponding with the galactophore image is as one
Group training parameter；
By the multiple galactophore image respectively corresponding one group of training parameter to the initial mammary gland parting model into
Row training, with the reference parameter of the determination initial mammary gland parting model.
Above-mentioned computer equipment 90 is for detecting in galactophore image during lump, and processor 910 is by calling storage
The operational order (operational order is storable in operating system) that device 940 stores executes following process:
Obtain target biological tissue image；
Determine the parting type of the target biological tissue image；
Determine that target object and the target object in the target biological tissue image are the probability of lump；
It will be general corresponding to parting type of the probability with the target biological tissue image that the target object is lump
Rate threshold value is compared, and probability threshold value corresponding to the biological tissue images of different parting types is different；
If the probability that the target object is lump is greater than corresponding to the parting type of the target biological tissue image
Probability threshold value then detects that the target object is lump.
Scheme provided by the embodiments of the present application can detecte out biological tissue images when detecting biological tissue images
Parting type, there is different lump probability threshold values in the biological tissue of different parting types, to improve the accurate of Mass detection
Degree.
In a kind of possible implementation, processor 910 is used for: when biological tissue images are galactophore image, passing through mesh
Mark mammary gland parting model determines the parting type of target galactophore image, and the target mammary gland parting model is by different parting classes
What multiple galactophore images of type and the parting type information of each galactophore image were trained.
In a kind of possible implementation, processor 910 is used for: marking the lump being detected；
Input and output (I/O) interface 930 is used for: output is comprising with markd galactophore image.
In a kind of possible implementation, the difference parting types include at least two in following parting type: rouge
Fat type, few body of gland type, polyadenous figure and dense form；
The corresponding probability threshold value of the lard type is a, few corresponding probability threshold value of body of gland type is b, the polyadenous body
The corresponding probability threshold value of type is c, and the corresponding probability threshold value of the dense form is d, and described a, b, c and d are both greater than 0, and a <b < c <
d。
In a kind of possible implementation, processor 910 is used for:
The target galactophore image is pre-processed and divided, divides subgraph to determine；
The segmentation subgraph is inputted into disaggregated model, determines target sub-object included in each segmentation subgraph,
The target sub-object is contained in the target object；
Determine that the target object is the probability of lump according to each target sub-object.
In a kind of possible implementation, processor 910 is used for:
Merge region overlapped in each target sub-object；
According to the set of the target sub-object after merging, determine that the target object is the probability of lump.
On to computer equipment 90 description can the description refering to fig. 1 to the part Fig. 6 understand that this place is not repeated
It repeats.
In the above-described embodiments, can come wholly or partly by software, hardware, firmware or any combination thereof real
It is existing.When implemented in software, it can entirely or partly realize in the form of a computer program product.
The computer program product includes one or more computer instructions.Load and execute on computers the meter
When calculation machine program instruction, entirely or partly generate according to process or function described in the embodiment of the present application.The computer can
To be general purpose computer, special purpose computer, computer network or other programmable devices.The computer instruction can be deposited
Storage in a computer-readable storage medium, or from a computer readable storage medium to another computer readable storage medium
Transmission, for example, the computer instruction can pass through wired (example from a web-site, computer, server or data center
Such as coaxial cable, optical fiber, Digital Subscriber Line (DSL)) or wireless (such as infrared, wireless, microwave) mode to another website
Website, computer, server or data center are transmitted.The computer readable storage medium can be computer and can deposit
Any usable medium of storage either includes that the data storages such as one or more usable mediums integrated server, data center are set
It is standby.The usable medium can be magnetic medium, (for example, floppy disk, hard disk, tape), optical medium (for example, DVD) or partly lead
Body medium (such as solid state hard disk Solid State Disk (SSD)) etc..
Those of ordinary skill in the art will appreciate that all or part of the steps in the various methods of above-described embodiment is can
It is completed with instructing relevant hardware by program, which can be stored in a computer readable storage medium, storage
Medium may include: ROM, RAM, disk or CD etc..
Above to the method, the method for training pattern, dress of lump in detection galactophore image provided by the embodiment of the present application
It sets and storage medium is described in detail, specific case used herein carries out the principle and embodiment of the application
It illustrates, the description of the example is only used to help understand the method for the present application and its core ideas；Meanwhile for ability
The those skilled in the art in domain, according to the thought of the application, there will be changes in the specific implementation manner and application range, comprehensive
Upper described, the contents of this specification should not be construed as limiting the present application.
Claims (15)
1. a kind of method of lump in detection biological tissue images characterized by comprising
Obtain target biological tissue image；
Determine the parting type of the target biological tissue image；
Determine that target object and the target object in the target biological tissue image are the probability of lump；
By probability threshold corresponding to the parting type of probability and the target biological tissue image that the target object is lump
Value is compared, and probability threshold value corresponding to the biological tissue images of different parting types is different；
If the probability that the target object is lump is greater than probability corresponding to the parting type of the target biological tissue image
Threshold value then detects that the target object is lump.
2. the method according to claim 1, wherein the biological tissue images are galactophore image, the determination
The parting type of the target biological tissue image, comprising:
The parting type of target galactophore image is determined by target mammary gland parting model, the target mammary gland parting model is to pass through
What multiple galactophore images of different parting types and the parting type information of each galactophore image were trained.
3. according to the method described in claim 2, it is characterized in that, the method also includes:
Mark the lump being detected；
Output is comprising with markd galactophore image.
4. according to the method in claim 2 or 3, which is characterized in that the difference parting type includes following parting type
In at least two: lard type, few body of gland type, polyadenous figure and dense form；
The corresponding probability threshold value of the lard type is a, few corresponding probability threshold value of body of gland type is b, the polyadenous figure pair
The probability threshold value answered is c, and the corresponding probability threshold value of the dense form is d, and described a, b, c and d are both greater than 0, and a <b < c < d.
5. according to the method in claim 2 or 3, which is characterized in that in the determination target biological tissue image
Target object and the target object are the probability of lump, comprising:
The target galactophore image is pre-processed and divided, divides subgraph to determine；
The segmentation subgraph is inputted into disaggregated model, determines target sub-object included in each segmentation subgraph, it is described
Target sub-object is contained in the target object；
Determine that the target object is the probability of lump according to each target sub-object.
6. according to the method described in claim 5, it is characterized in that, described determine the target pair according to each target sub-object
As the probability for lump, comprising:
Merge region overlapped in each target sub-object；
According to the set of the target sub-object after merging, determine that the target object is the probability of lump.
7. a kind of method of training pattern characterized by comprising
Sample image set is obtained, the sample image set includes multiple biological tissue images of different parting types, and
The parting type information of each biological tissue images；
Initial biological tissue's parting model is trained by described image set, with determination initial biological tissue's parting
The reference parameter of model；
The reference parameter is inputted into initial biological tissue's parting model, is used for target biological tissue parting mould to determine
Type, the target biological tissue parting model are used to determine the parting type of biological tissue images.
8. the method according to the description of claim 7 is characterized in that the biological tissue images be galactophore image, it is described to pass through
Described image set is defeated to be trained initial biological tissue's parting model, with determination initial biological tissue's parting model
Reference parameter, comprising:
Extract the characteristic information of each galactophore image；
By the characteristic information of each galactophore image, and parting type information corresponding with the galactophore image is as one group of instruction
Practice parameter；
By multiple galactophore images, respectively corresponding one group of training parameter is trained the initial mammary gland parting model, with
Determine the reference parameter of the initial mammary gland parting model.
9. the device of lump in a kind of detection galactophore image characterized by comprising
Acquiring unit, for obtaining target biological tissue image；
First determination unit, the parting type of the target biological tissue image for determining the acquiring unit acquisition；
Second determination unit, the target object in target biological tissue image for determining the acquiring unit acquisition, and
The target object is the probability of lump；
Comparing unit, the target object for determining second determination unit are that the probability of lump and described first determine list
Probability threshold value corresponding to the parting type for the target biological tissue image that member determines is compared, the biology of different parting types
Probability threshold value corresponding to organization chart picture is different；
Detection unit, if comparing the probability that target object is lump for the comparing unit is greater than the target biological tissue
When probability threshold value corresponding to the parting type of image, then detect that the target object is lump.
10. device according to claim 9, which is characterized in that
First determination unit, for passing through target mammary gland parting model when the biological tissue images are galactophore image
Determine the parting type of target galactophore image, the target mammary gland parting model is multiple mammary gland figures by different parting types
What the parting type information of picture and each galactophore image was trained.
11. device according to claim 10, which is characterized in that described device further include:
Marking unit, for marking by lump detected by the detecting unit；
Output unit includes the galactophore image with marking unit label for exporting.
12. a kind of device of training pattern characterized by comprising
Acquiring unit, for obtaining sample image set, the sample image set includes multiple biologies of different parting types
The parting type information of organization chart picture and each biological tissue images；
Training unit, the image collection for being obtained by the acquiring unit instruct initial biological tissue's parting model
Practice, with the reference parameter of determination initial biological tissue's parting model；
Determination unit, the reference parameter for training the training unit input initial biological tissue's parting model,
It is used for target biological tissue parting model to determine, the target biological tissue parting model to be for determining biological tissue images
Parting type.
13. device according to claim 12, which is characterized in that
The training unit is used for: when the biological tissue images are galactophore image,
Extract the characteristic information of each galactophore image；
By the characteristic information of each galactophore image, and parting type information corresponding with the galactophore image is as one group of instruction
Practice parameter；
By multiple galactophore images, respectively corresponding one group of training parameter is trained the initial mammary gland parting model, with
Determine the reference parameter of the initial mammary gland parting model.
14. a kind of computer equipment, which is characterized in that the computer equipment includes: input/output (I/O) interface, processor
And memory, program instruction is stored in the memory；
The processor is used to execute the program instruction stored in memory, executes the method as described in claim 1-6 is any,
Alternatively, any method of claim 7-8.
15. a kind of medical image detection system, which is characterized in that the medical image detection system include image scanning apparatus and
Image processing equipment；
Described image scanning device sends the medical image for scanning medical image, and to described image processing equipment；
Image processing equipment requires any method of 1-6 for perform claim.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910684650.5A CN110428405A (en) | 2019-01-30 | 2019-01-30 | Method, relevant device and the medium of lump in a kind of detection biological tissue images |
| CN201910092954.2A CN109872307B (en) | 2019-01-30 | 2019-01-30 | Method for detecting tumor in biological tissue image, corresponding device and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910092954.2A CN109872307B (en) | 2019-01-30 | 2019-01-30 | Method for detecting tumor in biological tissue image, corresponding device and medium |
Related Child Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910684650.5A Division CN110428405A (en) | 2019-01-30 | 2019-01-30 | Method, relevant device and the medium of lump in a kind of detection biological tissue images |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109872307A true CN109872307A (en) | 2019-06-11 |
| CN109872307B CN109872307B (en) | 2022-01-07 |
Family
ID=66918282
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910684650.5A Pending CN110428405A (en) | 2019-01-30 | 2019-01-30 | Method, relevant device and the medium of lump in a kind of detection biological tissue images |
| CN201910092954.2A Active CN109872307B (en) | 2019-01-30 | 2019-01-30 | Method for detecting tumor in biological tissue image, corresponding device and medium |
Family Applications Before (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910684650.5A Pending CN110428405A (en) | 2019-01-30 | 2019-01-30 | Method, relevant device and the medium of lump in a kind of detection biological tissue images |
Country Status (1)
| Country | Link |
|---|---|
| CN (2) | CN110428405A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110414539A (en) * | 2019-08-05 | 2019-11-05 | 腾讯科技（深圳）有限公司 | A kind of method and relevant apparatus for extracting characterization information |
| CN111080642A (en) * | 2019-12-31 | 2020-04-28 | 北京推想科技有限公司 | Tissue typing method and device based on medical image and electronic equipment |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101401730A (en) * | 2008-11-14 | 2009-04-08 | 南京大学 | Fast detecting method for mammary gland tumour shadiness area based on layered structure |
| CN104732213A (en) * | 2015-03-23 | 2015-06-24 | 中山大学 | Computer-assisted lump detecting method based on mammary gland magnetic resonance image |
| CN104771228A (en) * | 2015-03-23 | 2015-07-15 | 中山大学 | Method and device for judging whether breast mass is benign or malignant |
| CN106355043A (en) * | 2015-11-19 | 2017-01-25 | 小红象医疗科技有限公司 | Automatic analysis system and method for breast infrared information |
| CN107958453A (en) * | 2017-12-01 | 2018-04-24 | 深圳蓝韵医学影像有限公司 | Detection method, device and the computer-readable storage medium of galactophore image lesion region |
| CN108395986A (en) * | 2018-03-23 | 2018-08-14 | 余晖 | Human papilloma virus automatic parting direction detection device based on deep learning |
| CN108464840A (en) * | 2017-12-26 | 2018-08-31 | 安徽科大讯飞医疗信息技术有限公司 | A kind of breast lump automatic testing method and system |
| WO2018192672A1 (en) * | 2017-04-19 | 2018-10-25 | Siemens Healthcare Gmbh | Target detection in latent space |
| CN109191424A (en) * | 2018-07-23 | 2019-01-11 | 哈尔滨工业大学（深圳） | A kind of detection of breast lump and categorizing system, computer readable storage medium |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9392986B2 (en) * | 2011-02-14 | 2016-07-19 | University Of Rochester | Method and apparatus for cone beam breast CT image-based computer-aided detection and diagnosis |
| US20180306794A1 (en) * | 2017-04-20 | 2018-10-25 | The Regents Of The University Of California | Methods of Producing Gene Expression Profiles of Subjects Having Cancer and Kits for Practicing Same |
| CN108921821A (en) * | 2018-06-01 | 2018-11-30 | 中国人民解放军战略支援部队信息工程大学 | Method of discrimination based on the LASSO mammary cancer armpit lymph gland transfering state returned |
| CN109146848A (en) * | 2018-07-23 | 2019-01-04 | 东北大学 | A kind of area of computer aided frame of reference and method merging multi-modal galactophore image |
-
2019
- 2019-01-30 CN CN201910684650.5A patent/CN110428405A/en active Pending
- 2019-01-30 CN CN201910092954.2A patent/CN109872307B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101401730A (en) * | 2008-11-14 | 2009-04-08 | 南京大学 | Fast detecting method for mammary gland tumour shadiness area based on layered structure |
| CN104732213A (en) * | 2015-03-23 | 2015-06-24 | 中山大学 | Computer-assisted lump detecting method based on mammary gland magnetic resonance image |
| CN104771228A (en) * | 2015-03-23 | 2015-07-15 | 中山大学 | Method and device for judging whether breast mass is benign or malignant |
| CN106355043A (en) * | 2015-11-19 | 2017-01-25 | 小红象医疗科技有限公司 | Automatic analysis system and method for breast infrared information |
| WO2018192672A1 (en) * | 2017-04-19 | 2018-10-25 | Siemens Healthcare Gmbh | Target detection in latent space |
| CN107958453A (en) * | 2017-12-01 | 2018-04-24 | 深圳蓝韵医学影像有限公司 | Detection method, device and the computer-readable storage medium of galactophore image lesion region |
| CN108464840A (en) * | 2017-12-26 | 2018-08-31 | 安徽科大讯飞医疗信息技术有限公司 | A kind of breast lump automatic testing method and system |
| CN108395986A (en) * | 2018-03-23 | 2018-08-14 | 余晖 | Human papilloma virus automatic parting direction detection device based on deep learning |
| CN109191424A (en) * | 2018-07-23 | 2019-01-11 | 哈尔滨工业大学（深圳） | A kind of detection of breast lump and categorizing system, computer readable storage medium |
Non-Patent Citations (4)
| Title |
|---|
| TSUTSUMI, T等: "Total-length invasion of the facial nerve by parotid carcinoma ex pleomorphic adenoma", 《AURIS NASUS LARYNX》 * |
| 冯友红: "X线钼靶在早期乳腺癌诊断中的应用", 《临床合理用药》 * |
| 岳学旺 等: "《实用医学影像学上》", 30 September 2017 * |
| 马慧彬: "《基于机器学习的乳腺图像辅助诊断算法研究》", 31 August 2016 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110414539A (en) * | 2019-08-05 | 2019-11-05 | 腾讯科技（深圳）有限公司 | A kind of method and relevant apparatus for extracting characterization information |
| CN111080642A (en) * | 2019-12-31 | 2020-04-28 | 北京推想科技有限公司 | Tissue typing method and device based on medical image and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109872307B (en) | 2022-01-07 |
| CN110428405A (en) | 2019-11-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Adal et al. | Automated detection of microaneurysms using scale-adapted blob analysis and semi-supervised learning | |
| Ramteke et al. | Automatic medical image classification and abnormality detection using k-nearest neighbour | |
| Oliver et al. | A review of automatic mass detection and segmentation in mammographic images | |
| US8175351B2 (en) | Computer-aided detection and classification of suspicious masses in breast imagery | |
| Srinivas et al. | Unsupervised learning algorithms for MRI brain tumor segmentation | |
| CN109635846A (en) | A kind of multiclass medical image judgment method and system | |
| Ben-Ari et al. | Domain specific convolutional neural nets for detection of architectural distortion in mammograms | |
| WO2019184851A1 (en) | Image processing method and apparatus, and training method for neural network model | |
| CN110729045A (en) | Tongue image segmentation method based on context-aware residual error network | |
| CN109614869A (en) | A kind of pathological image classification method based on multi-scale compress rewards and punishments network | |
| Rani et al. | Superpixel with nanoscale imaging and boosted deep convolutional neural network concept for lung tumor classification | |
| Deepak et al. | Detection and discrimination of disease-related abnormalities based on learning normal cases | |
| Salazar-Licea et al. | Location of mammograms ROI's and reduction of false-positive | |
| Debnath et al. | Brain tumour segmentation using memory based learning method | |
| CN109872307A (en) | Method, relevant device and the medium of lump in a kind of detection biological tissue images | |
| Zhang et al. | A computer aided diagnosis system in mammography using artificial neural networks | |
| CN113177554A (en) | Thyroid nodule identification and segmentation method, system, storage medium and equipment | |
| Moradi et al. | Multi-class segmentation of skin lesions via joint dictionary learning | |
| Sharin et al. | Prostate cancer classification technique on pelvis CT images | |
| Yuan et al. | Automatic construction of filter tree by genetic programming for ultrasound guidance image segmentation | |
| Wang et al. | A r-cnn based approach for microaneurysm detection in retinal fundus images | |
| Vijayarajan et al. | A novel comparative study on breast cancer detection using different types of classification techniques | |
| Peter et al. | Medical image analysis using unsupervised and supervised classification techniques | |
| Jyothula et al. | Lung Cancer Detection in CT Scans Employing Image Processing Techniques and Classification by Decision Tree (DT) and K-Nearest Neighbor (KNN) | |
| Awang et al. | An overview of segmentation and classification techniques: A survey of brain tumour-related research |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |