CN109191424A - A kind of detection of breast lump and categorizing system, computer readable storage medium - Google Patents
A kind of detection of breast lump and categorizing system, computer readable storage medium Download PDFInfo
- Publication number
- CN109191424A CN109191424A CN201810810971.0A CN201810810971A CN109191424A CN 109191424 A CN109191424 A CN 109191424A CN 201810810971 A CN201810810971 A CN 201810810971A CN 109191424 A CN109191424 A CN 109191424A
- Authority
- CN
- China
- Prior art keywords
- pixel
- galactophore image
- detection
- galactophore
- breast
- 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
- 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
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—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
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Apparatus For Radiation Diagnosis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of detections of breast lump and categorizing system, computer readable storage medium, classified by each pixel to galactophore image, directly pixel is divided into three classes: generic pixel, benign tumors pixel and Malignant mass pixel, it realizes and Mass detection and lump classification rapidly and accurately is carried out to galactophore image, the technical issues of overcoming the prior art to isolate two steps of Mass detection and classification, causing accuracy rate low and low efficiency.
Description
Technical field
The present invention relates to medical domain, especially a kind of breast lump detection and categorizing system, computer-readable storage medium
Matter.
Background technique
Nipple correction image is widely used breast cancer image, has at low cost, quality height and high sexual valence
Compare the advantages of.Such image mainly reflects the isopycnic of breast tissue.Using nipple correction image, doctor can be with
It preferably determines breast lump and lump property is judged.But there is rely on subjective warp for doctor's artificial cognition
It tests, and the differentiation result different problems between different doctors.Since computer vision understands the development of technology, breast lump
And its qualitative computer is sentenced knowledge automatically and is possibly realized.
The existing breast lump detection of the overwhelming majority is divided into two steps of breast lump detection and classification with knowledge method is sentenced.Mesh
What front method was implemented in two steps is primarily due to as follows: it has been recognized that breast lump has relatively special texture structure,
Lump shortage compares clearly edge, therefore breast lump detection is one and distinguishes biggish with general goals Detection task
Business.Textural characteristics generally have outstanding feature in common object edge.Therefore, people tend to examine using special method
Classify again after measuring lump.Such processing method haves the shortcomings that computation complexity is high.Moreover, the two independent steps are cut
The inner link between breast lump detection and the benign task that the two are closely connected of classifying with Malignant mass is split；The two
Step respectively using but also the overall performance of method is limited, if the accuracy rate of the two steps is 80%, then
Final accuracy rate only has 64% on the whole.Therefore, it is badly in need of exploring new method and technology path, overcomes lacking for current method
It falls into.
Summary of the invention
The present invention is directed to solve at least some of the technical problems in related technologies.For this purpose, of the invention
One purpose is to provide a kind of detection of breast lump and categorizing system, computer readable storage medium, for realizing breast lump
Detection and classification.
The technical scheme adopted by the invention is that: a kind of detection of breast lump and categorizing system, including
Image acquisition unit, for obtaining galactophore image；
Pixel classifications unit is classified for each pixel to the galactophore image, the classification packet of the pixel
Include generic pixel, benign tumors pixel and Malignant mass pixel；
Detection and taxon, for carrying out breast lump detection and classification to the galactophore image after pixel classifications.
Further, the breast lump detection and categorizing system further include neural network unit, the neural network list
Member includes:
Network struction module, for constructing deep neural network, the deep neural network is used for the galactophore image
Each pixel classify, the classification of the pixel includes generic pixel, benign tumors pixel and Malignant mass pixel；Institute
It states deep neural network and is divided into six layers, be followed successively by the first convolutional layer, the second convolutional layer, third convolutional layer, first in sequence entirely
Articulamentum, the second full articulamentum and network output layer, each layer of the neuron number and mammary gland figure of the deep neural network
The number of pixels of picture is identical；
Network training module, for utilizing the galactophore image sample set training deep neural network.
Further, first convolutional layer, the second convolutional layer, third convolutional layer and the first full articulamentum are using improved
LU activation primitive will be described when the improved LU activation primitive is that the calculated value of LU activation primitive is less than or equal to preset value
The calculated value of LU activation primitive is set as 0.
Further, the described second full articulamentum uses sigmoid activation primitive.
Further, the galactophore image passes through first convolutional layer, the second convolutional layer, third convolutional layer, first entirely
The scalar of corresponding each pixel is calculated in articulamentum and the second full articulamentum；
The network output layer is used to calculate the absolute value of the difference of the scalar and pixel number category, the pixel number
Word category respectively includes generic pixel category, benign tumors pixel category and Malignant mass pixel category, and absolute value is minimum
Difference corresponding to pixel number category exported as the category of pixel.
Further, the generic pixel class is designated as -1, and the benign tumors pixel class is designated as 0, the Malignant mass picture
Plain class is designated as 1.
Further, the network training module includes
Image procossing submodule, for handling the galactophore image sample set, the galactophore image sample set packet
Multiple galactophore image samples are included, the method for handling the galactophore image sample set includes:
It carries out grey scale pixel value to the galactophore image sample to convert, including the pixel to the galactophore image sample
It carries out increasing gray value in proportion or gray value is reduced in proportion to the pixel of the galactophore image sample；
And/or
Gaussian noise is added to the pixel of all or part of galactophore image sample；
And/or
Salt-pepper noise is added to the pixel of all or part of galactophore image sample；
Training submodule, for utilizing described image processing submodule treated the galactophore image sample set training depth
Spend neural network.
Further, described detect with taxon includes
Region division module, for the galactophore image after pixel classifications to be divided into the subregion of multiple portions overlapping, institute
State the of different sizes of subregion；
Mass detection module, for judging that the benign tumors pixel of the subregion accounts for the ratio of total pixel of the subregion
Whether the ratio for total pixel that example or Malignant mass pixel account for the subregion is greater than preset ratio, when the judgment result is yes,
Judge the subregion for breast lump；
Lump categorization module, for judge the benign tumors pixel account for total pixel of the subregion ratio it is whether big
The ratio of total pixel of the subregion is accounted in Malignant mass pixel, if the determination result is YES, then the subregion is benign cream
Adenoncus block, conversely, the subregion is malignant breast tumors.
Further, the preset ratio is 0.3.
It is of the present invention another solution is that a kind of computer readable storage medium, is stored thereon with computer
Program, the computer program perform the steps of when being executed by processor
Obtain galactophore image；
Classify to each pixel of the galactophore image, the classification of the pixel includes generic pixel, benign swollen
Block pixel and Malignant mass pixel；
Breast lump detection and classification are carried out to the galactophore image after pixel classifications.
The beneficial effects of the present invention are:
A kind of breast lump detection of the present invention and categorizing system, computer readable storage medium, by galactophore image
Each pixel is classified, and directly pixel is divided into three classes: generic pixel, benign tumors pixel and Malignant mass pixel, real
Mass detection rapidly and accurately now is carried out to galactophore image and lump is classified, the prior art is overcome to isolate Mass detection and classification two
A step, the technical issues of causing accuracy rate low and low efficiency.
Detailed description of the invention
Specific embodiments of the present invention will be further explained with reference to the accompanying drawing:
Fig. 1 is an a kind of specific embodiment structural schematic diagram of breast lump detection and categorizing system in the present invention；
Fig. 2 is a specific embodiment function curve diagram of improved LU activation primitive in the present invention；
Fig. 3 is the specific embodiment schematic diagram for carrying out increasing gray value in proportion to pixel in the present invention；
Fig. 4 is the specific embodiment schematic diagram for carrying out reducing gray value in proportion to pixel in the present invention.
Specific embodiment
It should be noted that in the absence of conflict, the features in the embodiments and the embodiments of the present application can phase
Mutually combination.
A kind of detection of breast lump and categorizing system, with reference to Fig. 1, Fig. 1 be in the present invention a kind of breast lump detection and point
One specific embodiment structural schematic diagram of class system；Include:
Image acquisition unit, for obtaining galactophore image；In the present embodiment, galactophore image is to be obtained using x-ray instrument
Nipple correction image, original nipple correction image may be of different sizes, therefore, galactophore image of the present invention
The size of 200 pixel *, 200 pixel is zoomed to, pixel classifications unit is then inputted.
Pixel classifications unit is classified for each pixel to galactophore image, and the classification of pixel includes common picture
Element, benign tumors pixel and Malignant mass pixel；
Detection and taxon, for carrying out breast lump detection and classification to the galactophore image after pixel classifications.This reality
It applies in example, image acquisition unit, pixel classifications unit and detection can be computer with taxon, and computer utilizes computer
Program realizes image acquisition, pixel classifications, Mass detection and classification.Image acquisition unit, which can directly acquire, is stored in advance in meter
Galactophore image in calculation machine or in real-time feeding computer, can also obtain the cream of its shooting by communicating with x-ray instrument
Gland image.
Pixel classifications unit in the present invention is directly based upon galactophore image, carries out pixel class to the single pixel of galactophore image
Do not differentiate, detection can differentiate that result orient breast lump simultaneously and sentence to lump type with taxon based on pixel in turn
Know result (benign breast lump or malignant breast tumors), realizes quickly and effectively breast lump detection and classification.
As the further improvement of technical solution, in the present embodiment, pixel classifications unit is to utilize deep neural network pair
Galactophore image carries out pixel classifications, therefore, before carrying out pixel classifications to galactophore image, needs first projected depth neural network.
Breast lump detection and categorizing system further include neural network unit, and neural network unit includes:
Network struction module, for constructing deep neural network, deep neural network is used for each to galactophore image
Pixel is classified, and the classification of pixel includes generic pixel, benign tumors pixel and Malignant mass pixel；Deep neural network point
It is six layers, is followed successively by the first convolutional layer, the second convolutional layer, third convolutional layer, the first full articulamentum, the second full connection in sequence
Layer and network output layer, each layer of neuron number of deep neural network are identical as the number of pixels of galactophore image；
Network training module, for utilizing galactophore image sample set training deep neural network.
Wherein, the first convolutional layer, the second convolutional layer, third convolutional layer and the first full articulamentum activate letter using improved LU
Number, when improved LU activation primitive is that the calculated value of LU activation primitive is less than or equal to preset value, by the calculating of LU activation primitive
Value is set as 0.Specifically be provided that if z > e (Z indicate current convolutional calculation as a result, e be greater than zero it is lesser often
Number), then f (z)=z, otherwise, f (z)=0.Assuming that e=1, then when z > e f (z)=z function curve as shown in Fig. 2, Fig. 2 is this
A specific embodiment function curve diagram of improved LU activation primitive in invention.In addition, output valve always big in network rise compared with
Big effect, and small output valve effect is smaller.And depth network enhances the method for network robustness commonly to block wherein one
Connection between a little neurons, artificially reaches the sparsity of connection.It blocks connection and the output valve of neuron is set as 0 actually
It is identical effect, but the latter is simpler, is increased without any calculation amount.The present invention proposes to enhance network sparsity
Above-mentioned improved LU activation primitive, makes network have the characteristics that structure is sparse, to enhance Mass detection in a straightforward manner
With the robustness of classification results.And the second full articulamentum uses sigmoid activation primitive.Sigmoid activation primitive, which can enhance, " to be divided
The nonlinear transformation ability of class method ".Different from traditional convolutional neural networks, the deep neural network in the present invention is not set
Pooling layers, primarily to reducing the loss of information.
Specifically, galactophore image is by the first convolutional layer, the second convolutional layer, third convolutional layer, the first full articulamentum and the
The scalar of corresponding each pixel is calculated in two full articulamentums；Network output layer is used to calculate the difference of scalar Yu pixel number category
The absolute value of value, pixel number category respectively include generic pixel category, benign tumors pixel category and Malignant mass pixel class
Mark, and exports using pixel number category corresponding to the smallest difference of absolute value as the category of pixel, according to scalar and
The similitude of three categories is quantified as thrin.In the present embodiment, generic pixel class is designated as -1, benign tumors pixel class
It is designated as 0, Malignant mass pixel class is designated as 1.The step of this makes breast lump detection classify with lump is combined into one, and can be simultaneously
The computational efficiency of training for promotion and application stage.In addition, the setting of -1,0,1 three category also complies between three pixel class
The directviewing description of difference, difference meets poor between generic pixel and benign tumors pixel less than difference between -1 and 1 between -1 and 0
It is different to be less than between generic pixel and Malignant mass pixel the fact that difference.For example, it is assumed that a pixel inputs neural network
It afterwards, is 3 in the scalar that the second full articulamentum finally exports, then network output layer calculates the absolute of the difference between 3 and -1,0,1
Value, and compare the size of 3 absolute values, select class of the corresponding category of that the smallest difference of absolute value as this pixel
Mark, in the present embodiment, it is the absolute value of the difference between 3 and 1 that absolute value is the smallest, then the class of this pixel is designated as 1.To sum up,
After galactophore image inputs neural network, in the classification of the respective pixel of network output layer output galactophore image；That is network output layer
The final output of neuron be -1,0 or 1, the respective pixel of the galactophore image of input is identified as generic pixel, benign respectively
Lump pixel or Malignant mass pixel.
Network training module pair is utilized after the completion of deep neural network building as the further improvement of technical solution
It is trained, so that the pixel classifications processing to galactophore image is better achieved.Specifically, network training module includes
Image procossing submodule, for handling galactophore image sample set, galactophore image sample set includes multiple creams
Gland image pattern, galactophore image sample have been labelled with lump and its classification in galactophore image, processing galactophore image sample set
Method includes:
Grey scale pixel value is carried out to galactophore image sample to convert, and is carried out including the pixel to galactophore image sample year-on-year
Example increases gray value or is reduced gray value in proportion to the pixel of galactophore image sample；
And/or
Gaussian noise is added to the pixel of all or part of galactophore image sample；
And/or
Salt-pepper noise is added to the pixel of all or part of galactophore image sample；
Training submodule, for utilizing image procossing submodule treated galactophore image sample set training depth nerve net
Network, by treated, galactophore image sample set input deep neural network carries out pixel classifications to realize the training to network.
In practical applications, the image for the same mammary gland that the x-ray instrument of different manufacturers obtains has larger difference, even
The image that the x-ray instrument of same model obtains also discrepant phenomenon, in addition, all producers can not be obtained in practical application
A large amount of photographs of breast molybdenum target instrument, it is also not possible to obtain the great amount of images sample of the breast molybdenum target instrument of same model.In order to
Enhancing system is to the adaptability of galactophore image difference, and the present invention is in the training stage using limited image sample, and generation is as far as possible
More galactophore image samples enhances the diversity of galactophore image sample set, since they can indicate different breast molybdenum target instrument
Imaging results, it is possible to " experience " of abundant neural network, mind of the enhancing training result to the adaptability of sample, after training
There to be preferable robustness through network.Processing modification is carried out by the pixel to galactophore image sample, to obtain more mammary gland
Image pattern；By the way that using the amending method of the modification of above-mentioned grey scale pixel value and noise addition, the sample of generation is more, effect simultaneously
Fruit is more preferable.Concrete mode is as follows:
1. the grey scale pixel value of pair galactophore image sample converts, respectively to pixel increase in proportion gray value with
Reduce gray value, in proportion to obtain new sample.In the mode for increasing gray value in proportion, if the modified value of pixel is big
In 255, then force to be set as 255.The formula for increasing gray value in proportion is a+bx, and a, b are that be greater than 1, x be preimage by coefficient and b
Element value.The result for increasing gray value in proportion is carried out to pixel as shown in figure 3, Fig. 3 is to carry out in proportion in the present invention to pixel
Increase a specific embodiment schematic diagram of gray value；Provided in Fig. 3 x from 0 be changed to 255 when, knot that pixel value increases in proportion
Fruit (as shown in y-coordinate).The formula for reducing gray value in proportion is c+dx, and c, d are coefficient and d is greater than 0 but is still former less than 1, x
Pixel value.The result for reducing gray value in proportion is carried out to pixel as shown in figure 4, Fig. 4 is to carry out on year-on-year basis in the present invention to pixel
Example reduces a specific embodiment schematic diagram of gray value；Provided in Fig. 4 x from 0 be changed to 255 when, what pixel value became smaller in proportion
As a result (as shown in y-coordinate).
2. the pixel of pair all galactophore image samples adds a degree of Gaussian noise.
3. the pixel of pair all galactophore image samples adds a degree of salt-pepper noise.
4. the pixel in the image of a pair galactophore image sample adds Gauss or salt-pepper noise at random.
Above-mentioned 4 schemes feature the changeability of sample from different perspectives, wherein the first string analog different model X
The systematic divergence of alpha cellulose a gage increases and reduces respectively pixel value two ways and the new samples obtained is made to have better representative
Property, and do not lose in two possible change directions biased.Second and third, the random errors of four scheme analog instruments, it is right
Difference between same model instrument has relatively good analog capability.This four schemes are used simultaneously, so that in the new samples generated
Pixel variation it is comprehensive preferably.To sum up, image procossing submodule modifies to original galactophore image pixel, to increase cream
The diversity of gland image pattern collection.
As the further improvement of technical solution, in actual use, after image acquisition unit obtains galactophore image, pixel point
Class unit carries out pixel classifications to galactophore image using trained neural network, obtains the pixel class of each corresponding pixel
Mark.Detection and taxon carry out breast lump detection and classification to the galactophore image after pixel classifications, specifically, detect and divide
Class unit includes:
Region division module, for the galactophore image after pixel classifications to be divided into the subregion of multiple portions overlapping, son
Region it is of different sizes, in the present embodiment, subregion is rectangular area, the size of rectangular area it is different (7 pixel of such as 7 pixel *, 9
9 pixel of pixel *, 11 pixel *, 11 pixel, 13 pixel *, 13 pixel)；Because actual breast lump be it is not of uniform size,
Detection and categorizing system of the invention judges subregion of different sizes, so that the breast lump size detected more accords with
It closes practical.If the galactophore image of input to be only divided into the subregion of non-overlap, the size of subregion is identical, then when mammary gland is swollen
When block is across two adjacent subregions, it is likely that missing inspection occur.
Mass detection module, for judging that the benign tumors pixel of subregion accounts for the ratio of total pixel of subregion or pernicious
Whether the ratio that lump pixel accounts for total pixel of subregion is greater than preset ratio, when the judgment result is yes, judges that subregion is
Breast lump；In the present embodiment, preset ratio 0.3.
Lump categorization module, for judge benign tumors pixel account for total pixel of subregion ratio whether be greater than it is pernicious swollen
Block pixel accounts for the ratio of total pixel of subregion, and if the determination result is YES, then subregion is benign breast lump, conversely, sub-district
Domain is malignant breast tumors.
Mass detection module and lump categorization module carry out Mass detection and classification to each sub-regions of galactophore image,
Mass detection and classification are directly carried out based on galactophore image pixel, it is efficiently convenient, and accuracy is high.
The present invention also provides a kind of computer readable storage mediums, are stored thereon with computer program, computer program quilt
Processor performs the steps of when executing
Obtain galactophore image；
Classify to each pixel of galactophore image, the classification of pixel include generic pixel, benign tumors pixel and
Malignant mass pixel；
Breast lump detection and classification are carried out to the galactophore image after pixel classifications.
A kind of course of work of the computer program of computer-readable recording medium storage please refers to above-mentioned breast lump inspection
The description with categorizing system is surveyed, is repeated no more.
In order to efficiently with easily realize based on nipple correction image carry out breast lump detection with sentence knowledge, no
It is same as conventional method, the present invention directly carries out the pixel of galactophore image to sentence knowledge, is determined as generic pixel, benign tumors picture
Three element, Malignant mass pixel classifications；Neural network exports the classification of the pixel of galactophore image；Mammary gland is obtained according to pixel category
Lump positioning and its for benign or malignant lump differentiation result.
It is to be illustrated to preferable implementation of the invention, but the invention is not limited to the implementation above
Example, those skilled in the art can also make various equivalent variations on the premise of without prejudice to spirit of the invention or replace
It changes, these equivalent deformations or replacement are all included in the scope defined by the claims of the present application.
Claims (10)
1. a kind of breast lump detection and categorizing system, which is characterized in that including
Image acquisition unit, for obtaining galactophore image；
Pixel classifications unit is classified for each pixel to the galactophore image, and the classification of the pixel includes general
Logical pixel, benign tumors pixel and Malignant mass pixel；
Detection and taxon, for carrying out breast lump detection and classification to the galactophore image after pixel classifications.
2. breast lump according to claim 1 detection and categorizing system, which is characterized in that the breast lump detection with
Categorizing system further includes neural network unit, and the neural network unit includes:
Network struction module, for constructing deep neural network, the deep neural network is used for the every of the galactophore image
One pixel is classified, and the classification of the pixel includes generic pixel, benign tumors pixel and Malignant mass pixel；The depth
Degree neural network is divided into six layers, is followed successively by the first convolutional layer, the second convolutional layer, third convolutional layer, the first full connection in sequence
Layer, the second full articulamentum and network output layer, each layer of neuron number of the deep neural network and galactophore image
Number of pixels is identical；
Network training module, for utilizing the galactophore image sample set training deep neural network.
3. breast lump according to claim 2 detection and categorizing system, which is characterized in that first convolutional layer, the
Two convolutional layers, third convolutional layer and the first full articulamentum use improved LU activation primitive, and the improved LU activation primitive is
When the calculated value of LU activation primitive is less than or equal to preset value, 0 is set by the calculated value of the LU activation primitive.
4. breast lump detection according to claim 2 and categorizing system, which is characterized in that the second full articulamentum is adopted
With sigmoid activation primitive.
5. according to the described in any item breast lump detections of claim 2 to 4 and categorizing system, which is characterized in that the mammary gland
Image is calculated by first convolutional layer, the second convolutional layer, third convolutional layer, the first full articulamentum and the second full articulamentum
To the scalar of each pixel of correspondence；
The network output layer is used to calculate the absolute value of the difference of the scalar and pixel number category, the pixel numeric class
Mark respectively includes generic pixel category, benign tumors pixel category and Malignant mass pixel category, and by the smallest difference of absolute value
The corresponding pixel number category of value is exported as the category of pixel.
6. breast lump detection according to claim 5 and categorizing system, which is characterized in that the generic pixel category
It is -1, the benign tumors pixel class is designated as 0, and the Malignant mass pixel class is designated as 1.
7. according to the described in any item breast lump detections of claim 2 to 4 and categorizing system, which is characterized in that the network
Training module includes
Image procossing submodule, for handling the galactophore image sample set, the galactophore image sample set includes more
A galactophore image sample, the method for handling the galactophore image sample set include:
Grey scale pixel value is carried out to the galactophore image sample to convert, and is carried out including the pixel to the galactophore image sample
Increase gray value in proportion or gray value is reduced in proportion to the pixel of the galactophore image sample；
And/or
Gaussian noise is added to the pixel of all or part of galactophore image sample；
And/or
Salt-pepper noise is added to the pixel of all or part of galactophore image sample；
Training submodule, for utilizing described image processing submodule treated the galactophore image sample set training depth mind
Through network.
8. breast lump detection according to any one of claims 1 to 4 and categorizing system, which is characterized in that the detection
Include with taxon
Region division module, for the galactophore image after pixel classifications to be divided into the subregion of multiple portions overlapping, the son
Region it is of different sizes；
Mass detection module, for judge the benign tumors pixel of the subregion account for total pixel of the subregion ratio or
Whether the ratio that Malignant mass pixel accounts for total pixel of the subregion is greater than preset ratio, when the judgment result is yes, judgement
The subregion is breast lump；
Whether lump categorization module is greater than evil for judging that the benign tumors pixel accounts for the ratio of total pixel of the subregion
Property lump pixel account for the subregion total pixel ratio, if the determination result is YES, then the subregion is swollen for benign breast
Block, conversely, the subregion is malignant breast tumors.
9. breast lump detection according to claim 8 and categorizing system, which is characterized in that the preset ratio is 0.3.
10. a kind of computer readable storage medium, which is characterized in that be stored thereon with computer program, the computer program
It is performed the steps of when being executed by processor
Obtain galactophore image；
Classify to each pixel of the galactophore image, the classification of the pixel includes generic pixel, benign tumors picture
Element and Malignant mass pixel；
Breast lump detection and classification are carried out to the galactophore image after pixel classifications.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810810971.0A CN109191424B (en) | 2018-07-23 | 2018-07-23 | Breast mass detection and classification system and computer-readable storage medium |
| PCT/CN2018/124655 WO2020019671A1 (en) | 2018-07-23 | 2018-12-28 | Breast lump detection and classification system and computer-readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810810971.0A CN109191424B (en) | 2018-07-23 | 2018-07-23 | Breast mass detection and classification system and computer-readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109191424A true CN109191424A (en) | 2019-01-11 |
| CN109191424B CN109191424B (en) | 2022-04-22 |
Family
ID=64937114
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810810971.0A Active CN109191424B (en) | 2018-07-23 | 2018-07-23 | Breast mass detection and classification system and computer-readable storage medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN109191424B (en) |
| WO (1) | WO2020019671A1 (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109785320A (en) * | 2019-01-28 | 2019-05-21 | 太原理工大学 | It is a kind of to be classified to mammograms and known method for distinguishing using modified AlexNet model |
| CN109872307A (en) * | 2019-01-30 | 2019-06-11 | 腾讯科技（深圳）有限公司 | Method, relevant device and the medium of lump in a kind of detection biological tissue images |
| CN110827296A (en) * | 2019-11-01 | 2020-02-21 | 南京信息工程大学 | Mammary gland X-ray image analysis method of multi-target integrated deep neural network |
| CN111767760A (en) * | 2019-04-01 | 2020-10-13 | 北京市商汤科技开发有限公司 | Living body detection method and apparatus, electronic device, and storage medium |
| WO2023060874A1 (en) * | 2021-10-12 | 2023-04-20 | 上海蜜度信息技术有限公司 | Picture classification and object detection synchronous processing method and system, storage medium, and terminal |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111554383B (en) * | 2020-04-24 | 2023-09-05 | 浙江杜比医疗科技有限公司 | Neural network for breast tumor detection and detection system thereof |
| CN112241954B (en) * | 2020-10-22 | 2024-03-15 | 上海海事大学 | Full-view self-adaptive segmentation network configuration method based on lump differentiation classification |
| CN112699948A (en) * | 2020-12-31 | 2021-04-23 | 无锡祥生医疗科技股份有限公司 | Ultrasonic breast lesion classification method and device and storage medium |
| CN115619641B (en) * | 2022-10-24 | 2023-06-02 | 中山大学附属第五医院 | FFDM-based breast image processing method, FFDM-based breast image processing system, FFDM-based terminal and FFDM-based breast image processing medium |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100002917A1 (en) * | 2006-06-29 | 2010-01-07 | Fujifilm Corporation | Medical image segmentation apparatus and medical image segmentation program |
| CN103747718A (en) * | 2012-03-21 | 2014-04-23 | 奥林巴斯医疗株式会社 | Image processing device |
| CN104408482A (en) * | 2014-12-08 | 2015-03-11 | 电子科技大学 | Detecting method for high-resolution SAR (Synthetic Aperture Radar) image object |
| CN104966100A (en) * | 2015-06-17 | 2015-10-07 | 北京交通大学 | A benign and malignant image lump classification method based on texture primitives |
| CN107464250A (en) * | 2017-07-03 | 2017-12-12 | 深圳市第二人民医院 | Tumor of breast automatic division method based on three-dimensional MRI image |
| CN107507195A (en) * | 2017-08-14 | 2017-12-22 | 四川大学 | The multi-modal nasopharyngeal carcinoma image partition methods of PET CT based on hypergraph model |
| CN107590806A (en) * | 2017-09-19 | 2018-01-16 | 陈烨 | A kind of detection method and system based on brain medical imaging |
| CN107845098A (en) * | 2017-11-14 | 2018-03-27 | 南京理工大学 | Liver cancer image full-automatic partition method based on random forest and fuzzy clustering |
| CN107886514A (en) * | 2017-11-22 | 2018-04-06 | 浙江中医药大学 | Breast molybdenum target image lump semantic segmentation method based on depth residual error network |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060116579A1 (en) * | 2004-11-29 | 2006-06-01 | Pai-Chi Li | Ultrasound imaging apparatus and method thereof |
| CN104851101A (en) * | 2015-05-25 | 2015-08-19 | 电子科技大学 | Brain tumor automatic segmentation method based on deep learning |
| CN107749061A (en) * | 2017-09-11 | 2018-03-02 | 天津大学 | Based on improved full convolutional neural networks brain tumor image partition method and device |
-
2018
- 2018-07-23 CN CN201810810971.0A patent/CN109191424B/en active Active
- 2018-12-28 WO PCT/CN2018/124655 patent/WO2020019671A1/en active Application Filing
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100002917A1 (en) * | 2006-06-29 | 2010-01-07 | Fujifilm Corporation | Medical image segmentation apparatus and medical image segmentation program |
| CN103747718A (en) * | 2012-03-21 | 2014-04-23 | 奥林巴斯医疗株式会社 | Image processing device |
| CN104408482A (en) * | 2014-12-08 | 2015-03-11 | 电子科技大学 | Detecting method for high-resolution SAR (Synthetic Aperture Radar) image object |
| CN104966100A (en) * | 2015-06-17 | 2015-10-07 | 北京交通大学 | A benign and malignant image lump classification method based on texture primitives |
| CN107464250A (en) * | 2017-07-03 | 2017-12-12 | 深圳市第二人民医院 | Tumor of breast automatic division method based on three-dimensional MRI image |
| CN107507195A (en) * | 2017-08-14 | 2017-12-22 | 四川大学 | The multi-modal nasopharyngeal carcinoma image partition methods of PET CT based on hypergraph model |
| CN107590806A (en) * | 2017-09-19 | 2018-01-16 | 陈烨 | A kind of detection method and system based on brain medical imaging |
| CN107845098A (en) * | 2017-11-14 | 2018-03-27 | 南京理工大学 | Liver cancer image full-automatic partition method based on random forest and fuzzy clustering |
| CN107886514A (en) * | 2017-11-22 | 2018-04-06 | 浙江中医药大学 | Breast molybdenum target image lump semantic segmentation method based on depth residual error network |
Non-Patent Citations (1)
| Title |
|---|
| 史盛君 等: "基于神经网络的乳腺X光片中肿块检测法", 《哈尔滨理工大学学报》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109785320A (en) * | 2019-01-28 | 2019-05-21 | 太原理工大学 | It is a kind of to be classified to mammograms and known method for distinguishing using modified AlexNet model |
| CN109872307A (en) * | 2019-01-30 | 2019-06-11 | 腾讯科技（深圳）有限公司 | Method, relevant device and the medium of lump in a kind of detection biological tissue images |
| CN109872307B (en) * | 2019-01-30 | 2022-01-07 | 腾讯科技（深圳）有限公司 | Method for detecting tumor in biological tissue image, corresponding device and medium |
| CN111767760A (en) * | 2019-04-01 | 2020-10-13 | 北京市商汤科技开发有限公司 | Living body detection method and apparatus, electronic device, and storage medium |
| CN110827296A (en) * | 2019-11-01 | 2020-02-21 | 南京信息工程大学 | Mammary gland X-ray image analysis method of multi-target integrated deep neural network |
| WO2023060874A1 (en) * | 2021-10-12 | 2023-04-20 | 上海蜜度信息技术有限公司 | Picture classification and object detection synchronous processing method and system, storage medium, and terminal |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109191424B (en) | 2022-04-22 |
| WO2020019671A1 (en) | 2020-01-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109191424A (en) | A kind of detection of breast lump and categorizing system, computer readable storage medium | |
| Gao et al. | Classification of CT brain images based on deep learning networks | |
| CN111401480B (en) | Novel mammary gland MRI automatic auxiliary diagnosis method based on fusion attention mechanism | |
| CN110766051A (en) | Lung nodule morphological classification method based on neural network | |
| CN109191476A (en) | The automatic segmentation of Biomedical Image based on U-net network structure | |
| CN108510004B (en) | Cell classification method and system based on deep residual error network | |
| Han et al. | Automated pathogenesis-based diagnosis of lumbar neural foraminal stenosis via deep multiscale multitask learning | |
| US9330336B2 (en) | Systems, methods, and media for on-line boosting of a classifier | |
| CN108898588A (en) | Therapeutic effect appraisal procedure based on time series, electronic equipment | |
| CN106780546B (en) | The personal identification method of motion blur encoded point based on convolutional neural networks | |
| CN108549912A (en) | A kind of medical image pulmonary nodule detection method based on machine learning | |
| CN111833322B (en) | Garbage multi-target detection method based on improved YOLOv3 | |
| CN112329871B (en) | Pulmonary nodule detection method based on self-correction convolution and channel attention mechanism | |
| CN107103334A (en) | It is a kind of based on the Lung neoplasm sorting technique of convolutional neural networks and dictionary to study | |
| CN112819063B (en) | Image identification method based on improved Focal loss function | |
| CN108764361A (en) | The operating mode's switch method of beam type oil pumping machine indicating diagram based on integrated study | |
| CN115546525A (en) | Multi-view clustering method and device, electronic equipment and storage medium | |
| CN107423771B (en) | Two-time-phase remote sensing image change detection method | |
| US20210256717A1 (en) | Edge-guided ranking loss for monocular depth prediction | |
| CN110910497B (en) | Method and system for realizing augmented reality map | |
| Turi et al. | Classification of Ethiopian coffee beans using imaging techniques | |
| Mikhaylichenko et al. | Automatic grading of knee osteoarthritis from plain radiographs using densely connected convolutional networks | |
| CN103279960B (en) | A kind of image partition method of human body cache based on X-ray backscatter images | |
| CN116524495A (en) | Traditional Chinese medicine microscopic identification method and system based on multidimensional channel attention mechanism | |
| CN115331021A (en) | Dynamic feature extraction and description method based on multilayer feature self-difference fusion |
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 |