CN109805963A - The determination method and system of one Endometrium parting - Google Patents
The determination method and system of one Endometrium parting Download PDFInfo
- Publication number
- CN109805963A CN109805963A CN201910223197.8A CN201910223197A CN109805963A CN 109805963 A CN109805963 A CN 109805963A CN 201910223197 A CN201910223197 A CN 201910223197A CN 109805963 A CN109805963 A CN 109805963A
- Authority
- CN
- China
- Prior art keywords
- endometrium
- image
- uterus
- standard
- parting
- 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
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses the determination methods and system of an Endometrium parting, this method comprises: obtaining uterus standard longitudal section image；Utilize endometrium parting model, the uterus standard longitudal section image is analyzed, obtain the endometrium genotyping result of the uterus standard longitudal section image, wherein the endometrium parting model is generated by having marked the uterus standard longitudal section image training of endometrium type.The endometrium parting model pre-established is utilized in the present invention, to the uterus, standard longitudal section image is analyzed, endometrium genotyping result can be automatically obtained, there is no need to doctors to carry out manual analysis, improve the precision and efficiency of the judgement of endometrium parting.
Description
Technical field
The present invention relates to technical field of information processing, more particularly to an Endometrium classifying method and system.
Background technique
Endometrium parting is to the success of IVF (in vitro fertilization, in vitro fertilization) and gamete in defeated ovum
Conveying in pipe is particularly significant.If in the IVF period on the day of oocyte pickup or the previous day endometrium is not A type, embryo
Implantation will not occur for tire or implantation rate is low.
The analyzing and diagnosing of endometrium at present, mainly doctor are according to the uterus longitudal section of the standard of acquisition, by existing
Knowledge carry out endometrium parting judgement.As it can be seen that the prior art depends on the experience of doctor, it is easy so that judgement essence
Degree reduces, and the working efficiency of artificial judgment is lower.
Summary of the invention
It is directed to the above problem, the present invention provides the determination method and system of an Endometrium parting, realizes raising
The precision and efficiency of endometrium parting judgement.
To achieve the goals above, the present invention provides the following technical scheme that
The determination method of one Endometrium parting, this method comprises:
Obtain uterus standard longitudal section image；
Using endometrium parting model, the uterus standard longitudal section image is analyzed, obtains the uterus mark
The endometrium genotyping result of quasi- longitudal section image, wherein the endometrium parting model is by having marked endometrium
What the uterus standard longitudal section image training of type generated.
Optionally, the acquisition uterus standard longitudal section image, comprising:
Obtain the ultrasonic scan image for being directed to uterus；
According to the characteristics of image in three layers of uterus in the ultrasonic scan image, uterus standard longitudal section image is determined.
Optionally, this method further include:
Based on the uterus standard longitudal section image for having marked endometrium type, endometrium parting model is created.
Optionally, described based on the uterus standard longitudal section image for having marked endometrium type, create endometrium point
Pattern type, comprising:
By the uterus standard longitudal section image for having marked endometrium type, it is determined as training sample；
Neural network model is set, the neural network model is detected based on the training sample and instruction of classifying
Practice, and be based on training result, the neural network model is adjusted, obtains endometrium parting model, wherein the mind
It is using being constructed using the DenseBlock structure in DenseNet network through network model.
Optionally, described to utilize endometrium parting model, the uterus standard longitudal section is analyzed, described in acquisition
The endometrium genotyping result of uterus standard longitudal section image, comprising:
Using the endometrium parting model, the uterus standard longitudal section is detected, obtains endometrium area
Domain；
Image characteristics extraction is carried out to the endometrium region, by the endometrium parting model to the extraction
Characteristics of image analyzed, obtain the endometrium genotyping result of the uterus standard longitudal section image.
Optionally, the setting neural network model examines the neural network model based on the training sample
Survey and classification based training, and based on detection training result, the neural network model is adjusted, endometrium parting mould is obtained
Type, comprising:
Build the multitask neural network of endometrium parting；
The training sample is inputted into the neural network, the operation of convolutional layer and pond layer is carried out to the training sample
Expand image channel number, obtain various sizes of deep layer characteristics of image, and the deep layer characteristics of image is separately input to described
The detection branches of neural network and classification branch；
By the training to the detection branches, detection block is obtained, the detection block is used to detect for endometrium
Image-region；
It is obtained by the training to the classification branch so that the image-region obtained to the detection block is classified
Endometrium type.
The decision-making system of one Endometrium parting, this method comprises:
Acquiring unit, for obtaining uterus standard longitudal section image；
Analytical unit is analyzed the uterus standard longitudal section image, is obtained for utilizing endometrium parting model
Obtain the endometrium genotyping result of the uterus standard longitudal section image, wherein the endometrium parting model is by
Mark the uterus standard longitudal section image training generation of endometrium type.
Optionally, the acquiring unit includes:
Image obtains subelement, for obtaining the ultrasonic scan image for being directed to uterus；
Image determines subelement, for the characteristics of image according to three layers of uterus in the ultrasonic scan image, determines son
Palace standard longitudal section image；
Wherein, the system further include:
Creating unit, for creating endometrium based on the uterus standard longitudal section image for having marked endometrium type
Parting model；
Wherein, the creating unit includes:
Subelement is determined, for being determined as instructing by the uterus standard longitudal section image for having marked endometrium type
Practice sample；
Training subelement, for neural network model to be arranged, based on the training sample to the neural network model into
Row detection and classification based training, and it is based on training result, the neural network model is adjusted, endometrium parting mould is obtained
Type, wherein the neural network model is to be constructed using the Dense Block structure in DenseNet network.
Optionally, the analytical unit includes:
Detection sub-unit, for being detected to the uterus standard longitudal section using the endometrium parting model,
Obtain endometrium region；
Parting subelement passes through the endometrium point for carrying out image characteristics extraction to the endometrium region
Pattern type analyzes the characteristics of image of the extraction, obtains the endometrium parting knot of the uterus standard longitudal section image
Fruit.
Optionally, the trained subelement includes:
Network establishment subelement, for building the multitask neural network of endometrium parting；
Image procossing subelement carries out the training sample for the training sample to be inputted the neural network
Image channel number is expanded in the operation of convolutional layer and pond layer, obtains various sizes of deep layer characteristics of image, and by the deep layer figure
As feature is separately input to detection branches and the classification branch of the neural network；
Training subelement is detected, for obtaining detection block by the training to the detection branches, the detection block is used for
Detection is directed to the image-region of endometrium；
Classification based training subelement, for by it is described classification branch training so as to the detection block obtain figure
As region is classified, endometrium type is obtained.
Compared to the prior art, it the present invention provides the determination method and system of an Endometrium parting, is obtaining
After the standard longitudal section image of uterus, using the endometrium parting model pre-established, to the uterus standard longitudal section image into
Row analysis, can automatically obtain endometrium genotyping result, and there is no need to doctors to carry out manual analysis, improve endometrium point
The precision and efficiency of type judgement.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
The embodiment of invention for those of ordinary skill in the art without creative efforts, can also basis
The attached drawing of offer obtains other attached drawings.
Fig. 1 is the flow diagram of the determination method of Endometrium parting provided in an embodiment of the present invention；
Fig. 2 is the schematic diagram of the network structure of Endometrium parting provided in an embodiment of the present invention；
Fig. 3 is the structural schematic diagram of the decision-making system of Endometrium parting provided in an embodiment of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
Term " first " and " second " in description and claims of this specification and above-mentioned attached drawing etc. are for area
Not different objects, rather than for describing specific sequence.Furthermore term " includes " and " having " and their any deformations,
It is intended to cover and non-exclusive includes.Such as it contains the process, method of a series of steps or units, system, product or sets
It is standby not to be set in listed step or unit, but may include the step of not listing or unit.
The determination method of an Endometrium parting is provided in embodiments of the present invention, referring to Fig. 1, this method comprises:
S101, uterus standard longitudal section image is obtained.
When obtaining uterus standard longitudal section image, can be realized according to following steps:
S1011, the ultrasonic scan image for being directed to uterus is obtained；
S1012, according to the characteristics of image in three layers of uterus in the ultrasonic scan image, determine uterus standard longitudal section figure
Picture.
Uterine wall is divided into three layers, and outer layer is serous coat, that is, perimetrium；Middle layer is strong thick muscle layer, i.e. mesometrium, by putting down
Sliding flesh composition；Internal layer is mucous membrane, i.e. endometrium.Uterine cavity line is shown centered in the ultrasonic scan image in uterus, the uterine cavity
Line is endometrium line, and endometrium is divided into mesometrium in both sides up and down.According to three layers of the characteristics of image in uterus, uterus is judged
Standard longitudal section obtains uterus standard longitudal section from a series of images of real-time scanning.
S102, using endometrium parting model, the uterus standard longitudal section image is analyzed, the son is obtained
The endometrium genotyping result of palace standard longitudal section image.
Wherein, endometrium parting mode is the uterus standard longitudal section image training by having marked endometrium type
It generates.It is not to use to pass through doctor in the prior art in embodiments of the present invention after obtaining uterus standard longitudal section image
Experience carries out parting judgement, but using the uterus standard longitudal section image as input information, it is input to the uterus being pre-created
In inner membrance parting model, feature extraction and analysis are carried out to image automatically by the model, to obtain endometrium genotyping result.
When carrying out parting judgement using endometrium parting model, which be may comprise steps of:
S1021, using endometrium parting model, the uterus standard longitudal section is detected, obtain endometrium
Region；
S1022, image characteristics extraction is carried out to the endometrium region, by the endometrium parting model to institute
The characteristics of image for stating extraction is analyzed, and the endometrium genotyping result of the uterus standard longitudal section image is obtained.
Uterus standard longitudal section image is input in the endometrium parting model, which can detect intrauterine
Film region, and according to the characteristics of image shown in the region, provide the affiliated type of endometrium in the image.Wherein,
Characteristics of image includes but is not limited to the grayscale information of image, marginal information etc., it usually needs is carried out to these features further
Abstract combination can just obtain the judgement of accurate type, image characteristics extraction and abstraction process are equal in embodiments of the present invention
For automation process.
The parting of endometrium includes A type, Type B and c-type.A type shows as a kind of multilayer " three lines " pattern endometrium, packet
Low echo inner membrance containing the stronger inner membrance outer layer of echo and uterine cavity middle line, and between the two；Type B is shown as in medium echo
Interbed inner membrance, the inner membrance outer layer of medium echo and the relatively low uterine cavity middle line of echo；C-type inner membrance, holostrome (including endometrium,
Inner membrance outer layer and uterine cavity middle line) medium echo is shown as, uterine cavity middle line is unclear.
Endometrium automatic parting direction algorithm uses deep learning network in embodiments of the present invention, belonging to automatic identification inner membrance
Type, therefore, this method in another embodiment of the invention further include:
S201, based on the uterus standard longitudal section image for having marked endometrium type, create endometrium parting model.
When creating endometrium parting model, which be may comprise steps of:
S2011, by the uterus standard longitudal section image for having marked endometrium type, be determined as training sample；
S2022, setting neural network model, are detected and are divided to the neural network model based on the training sample
Class training, and based on detection training result, the neural network model is adjusted, endometrium parting model is obtained.
Creation endometrium parting model is using deep learning nerual network technique in embodiments of the present invention, in order to
The precision that can be improved parting judgement, preferably by the Dense Block structure in DenseNet network, building is suitable for uterus
The deep learning network model of inner membrance automatic parting direction, i.e. endometrium parting model.
Under the premise of Dense Block guarantees that maximum information is transmitted between layers in network, will directly own
Layer connects, and this structure alleviates the phenomenon that gradient disappears in deep learning neural network training process, strengthens image
The transmitting of feature is more effectively multiplexed low layer pictures feature, is suitble to the processing to uterus standard longitudal section image.
In traditional convolutional neural networks, if network there are L layers, there is L connection, but the meeting in DenseNet network
There is L* (L+1)/2 connection, the output of each layer of input from all layers of front.Have benefited from the design of Dense Block, often
The characteristic image quantity of a convolutional layer output is seldom, it is possible to reduce DenseNet network parameter amount, this connection type but also
The transmitting of feature and gradient is more efficient, and the training of network is also more effective.
Specifically, the setting neural network model, examines the neural network model based on the training sample
Survey and classification based training, and based on detection training result, the neural network model is adjusted, endometrium parting mould is obtained
Type, comprising:
S301, the multitask neural network for building endometrium parting；
S302, the training sample is inputted into the neural network, convolutional layer and pond layer is carried out to the training sample
Operation expand image channel number, obtain various sizes of deep layer characteristics of image, and the deep layer characteristics of image is inputted respectively
Detection branches and classification branch to the neural network；
S303, by the training to the detection branches, obtain detection block, the detection block is directed to intrauterine for detecting
The image-region of film；
S304, by it is described classification branch training so as to the detection block obtain image-region classify,
Obtain endometrium type.
Wherein, by the training to the detection branches and the classification branch, the neural network that training is completed
The characteristics of image for being suitable for endometrium classification in image is enough extracted, endometrium type is obtained.
The multitask neural network for the endometrium parting built in above-mentioned steps S301 uses DenseNet network
In Dense Block structure.
Training sample is the uterus standard longitudal section image for having marked endometrium type, i.e., the training sample as
Data needed for network training, and the endometrium type in image can embody in tag form in the picture.It needs to unite
The image format of the one uterus standard longitudal section image of marking types.It can be according to intrauterine in the standard longitudal section image of uterus
Position, form and the feature of film mark endometrium region with rectangle frame, record the rectangle upper left corner and bottom right angular coordinate
Then value marks type described in endometrium, such as: A type inner membrance is labeled as 1, and Type B inner membrance is labeled as 2, and c-type inner membrance is labeled as
3, so that only one class label for determining of uterus standard longitudal section image.
The network structure for the endometrium parting being related in the technical solution provided by the present invention is as shown in Fig. 2, wherein wrap
Include detection and classification Liang Ge branch, the two branch's common network parts be the Dense Block for including in dotted line frame in Fig. 2 and
It is utilized in pond module, referred to as backbone, i.e. detection branches and the training process of classification branch by Dense Block
With the relevant image features of pondization treated input picture.It is specific:
In Fig. 2, training sample, that is, uterus standard longitudal section image is inputted, it is logical to first pass around a convolutional layer expansion image
Road number；Then various sizes of depth image feature in image is extracted using Dense Block and pondization, wherein pond is
Down-sampling, is used for downscaled images size, and pondization indicates that step-length is 2 2*2 maximum pond layer, that is, uses the sliding window of a 2*2
Mouth slides on the image, and step-length is each sliding length in pixels of window；It detects in output branch, has been multiplexed in backbone
Low layer pictures feature is up-sampled for characteristic pattern size to be gradually expanded, the rectangle frame of finally output detection endometrium, output figure
As identical as input image size；Dense Block further extracts characteristics of image and abstract combination in type output branch,
The probability value that input picture belongs to different type inner membrance is obtained by global mean value pond, inner membrance class is finally exported according to probability value
Type, i.e., using the maximum corresponding type of probability value as output result.
Network training detection simultaneously and two subtasks of classifying, mutually supervise and supplement between task, network is easier just
The characteristics of image in endometrium region is really extracted, detection and nicety of grading are improved.When training, detection and error in classification are anti-simultaneously
To propagation, the update of convolutional layer weight in backbone is supervised, neural network needs mass data to carry out successive ignition training, instruction
The characteristics of image that inner membrance classification is best suited in image can be automatically extracted by practicing the network completed, and be obtained by the analytical calculation to feature
To the endometrium position of prediction and genotyping result.
It should be noted that the Dense Block in Fig. 2 below Dense Block and pond module and in detection branches
With the printed words for being labeled with × 4 below up-sampling module, characterization is that above-mentioned two process can carry out successive ignition processing, figure
It is to have carried out 4 repetitive exercise processing to determine iteration as the case may be so that the result of training is more accurate in 2
Number, such as specific the number of iterations is determined according to loss function.
Therefore, the film type using film location in multitask network monitor and in judging in the present invention, multitask network can be
Better feature is obtained under limited data, multiple tasks are complementary to one another, the hiding feature being easy in discovery image, more to appoint
Business introduces supervision message more abundant from different aspect, to improve arithmetic accuracy, obtains more believable classification results.
Due to the information very abundant for including in the training sample image of input, therefrom mentioned automatically using deep learning network
Difference is smaller in taking-up class, then the biggish characteristics of image of difference between class makes correct categorised decision.Loss function is guiding net
The important link in network training direction, the loss function for being usually used in classification problem have least mean-square error, cross entropy etc..In the present invention
Single labeling problem in embodiment, can choose cross entropy as network losses function.For the non-linear table for improving model
Danone power can add activation primitive after each convolutional layer.In deep learning neural network training process, data set
The desired effect and robustness that quantity exports network training are all important factor in order, since the data volume of ultrasound image has
Limit carries out quantity enhancing so needing to use in the training process the methods of translate, cut out, overturn, increases sample size.Due to
It only needs to obtain endometrium genotyping result, therefore does not export the endometrium position detected, output category result.
The present invention provides the determination methods of an Endometrium parting, after obtaining uterus standard longitudal section image,
Using the endometrium parting model pre-established, to the uterus, standard longitudal section image is analyzed, and can automatically obtain son
Endometrium genotyping result, there is no need to doctors to carry out manual analysis, improve the precision and efficiency of the judgement of endometrium parting.
The decision-making system of an Endometrium parting is provided in another embodiment of the present invention, referring to Fig. 3, the system packet
It includes:
Acquiring unit 10, for obtaining uterus standard longitudal section image；
Analytical unit 20, for analyzing the uterus standard longitudal section image using endometrium parting model,
Obtain the endometrium genotyping result of the uterus standard longitudal section image, wherein the endometrium parting model is to pass through
The uterus standard longitudal section image training generation of endometrium type is marked.
The present invention provides the decision-making system of an Endometrium parting, acquiring unit is obtaining uterus standard longitudal section
After image, using the endometrium parting model pre-established in analytical unit, which is carried out
Analysis, can automatically obtain endometrium genotyping result, and there is no need to doctors to carry out manual analysis, improve endometrium parting
The precision and efficiency of judgement.
On the basis of the above embodiments, acquiring unit 10 includes:
Image obtains subelement, for obtaining the ultrasonic scan image for being directed to uterus；
Image determines subelement, for the characteristics of image according to three layers of uterus in the ultrasonic scan image, determines son
Palace standard longitudal section image.
On the basis of above-described embodiment, the decision-making system of the endometrium parting further include:
Creating unit, for creating endometrium based on the uterus standard longitudal section image for having marked endometrium type
Parting model.
On the basis of the above embodiments, the creating unit includes:
Subelement is determined, for being determined as instructing by the uterus standard longitudal section image for having marked endometrium type
Practice sample；
Training subelement, for neural network model to be arranged, based on the training sample to the neural network model into
Row detection and classification based training, and it is based on training result, the neural network model is adjusted, endometrium parting mould is obtained
Type, wherein the neural network model is using being constructed using the Dense Block structure in DenseNet network.
On the basis of the above embodiments, the analytical unit 20 in the system includes:
Detection sub-unit, for being detected to the uterus standard longitudal section using the endometrium parting model,
Obtain endometrium region；
Parting subelement passes through the endometrium point for carrying out image characteristics extraction to the endometrium region
Pattern type analyzes the characteristics of image of the extraction, obtains the endometrium parting knot of the uterus standard longitudal section image
Fruit.
In the above-described embodiments, the training subelement in creating unit includes:
Network establishment subelement, for building the multitask neural network of endometrium parting；
Image procossing subelement carries out the training sample for the training sample to be inputted the neural network
Image channel number is expanded in the operation of convolutional layer and pond layer, obtains various sizes of deep layer characteristics of image, and by the deep layer figure
As feature is separately input to detection branches and the classification branch of the neural network；
Training subelement is detected, for obtaining detection block by the training to the detection branches, the detection block is used for
Detection is directed to the image-region of endometrium；
Classification based training subelement, for by it is described classification branch training so as to the detection block obtain figure
As region is classified, endometrium type is obtained.
Each embodiment in this specification is described in a progressive manner, the highlights of each of the examples are with other
The difference of embodiment, the same or similar parts in each embodiment may refer to each other.For device disclosed in embodiment
For, since it is corresponded to the methods disclosed in the examples, so being described relatively simple, related place is said referring to method part
It is bright.
The foregoing description of the disclosed embodiments enables those skilled in the art to implement or use the present invention.
Various modifications to these embodiments will be readily apparent to those skilled in the art, as defined herein
General Principle can be realized in other embodiments without departing from the spirit or scope of the present invention.Therefore, of the invention
It is not intended to be limited to the embodiments shown herein, and is to fit to and the principles and novel features disclosed herein phase one
The widest scope of cause.
Claims (10)
1. the determination method of an Endometrium parting, which is characterized in that this method comprises:
Obtain uterus standard longitudal section image；
Using endometrium parting model, the uterus standard longitudal section image is analyzed, it is vertical to obtain the uterus standard
The endometrium genotyping result of tangent plane picture, wherein the endometrium parting model is by having marked endometrium type
Uterus standard longitudal section image training generate.
2. the method according to claim 1, wherein the acquisition uterus standard longitudal section image, comprising:
Obtain the ultrasonic scan image for being directed to uterus；
According to the characteristics of image in three layers of uterus in the ultrasonic scan image, uterus standard longitudal section image is determined.
3. the method according to claim 1, wherein this method further include:
Based on the uterus standard longitudal section image for having marked endometrium type, endometrium parting model is created.
4. according to the method described in claim 3, it is characterized in that, described based on the uterus standard for having marked endometrium type
Longitudal section image creates endometrium parting model, comprising:
By the uterus standard longitudal section image for having marked endometrium type, it is determined as training sample；
Neural network model is set, detection and classification based training are carried out to the neural network model based on the training sample, and
Based on training result, the neural network model is adjusted, obtains endometrium parting model, wherein the nerve net
Network model is to be constructed using the Dense Block structure in DenseNet network.
5. according to the method described in claim 4, it is characterized in that, described utilize endometrium parting model, to the uterus
Standard longitudal section is analyzed, and the endometrium genotyping result of the uterus standard longitudal section image is obtained, comprising:
Using the endometrium parting model, the uterus standard longitudal section is detected, obtains endometrium region；
Image characteristics extraction is carried out to the endometrium region, by the endometrium parting model to the figure of the extraction
As feature is analyzed, the endometrium genotyping result of the uterus standard longitudal section image is obtained.
6. according to the method described in claim 5, it is characterized in that, the setting neural network model, is based on the trained sample
This carries out detection and classification based training to the neural network model, and based on detection training result, to the neural network model
It is adjusted, obtains endometrium parting model, comprising:
Build the multitask neural network of endometrium parting；
The training sample is inputted into the neural network, the operation for carrying out convolutional layer and pond layer to the training sample is expanded
Image channel number obtains various sizes of deep layer characteristics of image, and the deep layer characteristics of image is separately input to the nerve
The detection branches of network and classification branch；
By the training to the detection branches, detection block is obtained, the detection block is used to detect the image for endometrium
Region；
Uterus is obtained so that the image-region obtained to the detection block is classified by the training to the classification branch
Interior film type.
7. the decision-making system of an Endometrium parting, which is characterized in that this method comprises:
Acquiring unit, for obtaining uterus standard longitudal section image；
Analytical unit analyzes the uterus standard longitudal section image, obtains institute for utilizing endometrium parting model
State the endometrium genotyping result of uterus standard longitudal section image, wherein the endometrium parting model is by having marked
What the uterus standard longitudal section image training of endometrium type generated.
8. system according to claim 7, which is characterized in that the acquiring unit includes:
Image obtains subelement, for obtaining the ultrasonic scan image for being directed to uterus；
Image determines subelement, for the characteristics of image according to three layers of uterus in the ultrasonic scan image, determines that uterus is marked
Quasi- longitudal section image；
Wherein, the system further include:
Creating unit, for creating endometrium parting based on the uterus standard longitudal section image for having marked endometrium type
Model；
Wherein, the creating unit includes:
It determines subelement, trains sample for being determined as the uterus standard longitudal section image for having marked endometrium type
This；
Training subelement examines the neural network model based on the training sample for neural network model to be arranged
Survey and classification based training, and it is based on training result, the neural network model is adjusted, endometrium parting model is obtained,
Wherein, the neural network model is using being constructed using the Dense Block structure in DenseNet network.
9. system according to claim 8, which is characterized in that the analytical unit includes:
Detection sub-unit detects the uterus standard longitudal section, obtains for utilizing the endometrium parting model
Endometrium region；
Parting subelement passes through the endometrium parting mould for carrying out image characteristics extraction to the endometrium region
Type analyzes the characteristics of image of the extraction, obtains the endometrium genotyping result of the uterus standard longitudal section image.
10. system according to claim 9, which is characterized in that the trained subelement includes:
Network establishment subelement, for building the multitask neural network of endometrium parting；
Image procossing subelement carries out convolution to the training sample for the training sample to be inputted the neural network
Image channel number is expanded in the operation of layer and pond layer, obtains various sizes of deep layer characteristics of image, and the deep layer image is special
Sign is separately input to detection branches and the classification branch of the neural network；
Training subelement is detected, for obtaining detection block, the detection block is for detecting by the training to the detection branches
For the image-region of endometrium；
Classification based training subelement, for by it is described classification branch training so as to the detection block obtain image district
Domain is classified, and endometrium type is obtained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910223197.8A CN109805963B (en) | 2019-03-22 | 2019-03-22 | Method and system for judging endometrium typing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910223197.8A CN109805963B (en) | 2019-03-22 | 2019-03-22 | Method and system for judging endometrium typing |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109805963A true CN109805963A (en) | 2019-05-28 |
| CN109805963B CN109805963B (en) | 2022-07-05 |
Family
ID=66610089
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910223197.8A Active CN109805963B (en) | 2019-03-22 | 2019-03-22 | Method and system for judging endometrium typing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109805963B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111436972A (en) * | 2020-04-13 | 2020-07-24 | 王时灿 | Three-dimensional ultrasonic gynecological disease diagnosis device |
| CN111768379A (en) * | 2020-06-29 | 2020-10-13 | 深圳度影医疗科技有限公司 | Standard section detection method of three-dimensional uterine ultrasound image |
| CN112215843A (en) * | 2019-12-31 | 2021-01-12 | 无锡祥生医疗科技股份有限公司 | Ultrasonic intelligent imaging navigation method and device, ultrasonic equipment and storage medium |
| CN112949723A (en) * | 2021-03-08 | 2021-06-11 | 西安美佳家医疗科技有限责任公司 | Endometrium pathology image classification method |
| CN113520317A (en) * | 2021-07-05 | 2021-10-22 | 汤姆飞思（香港）有限公司 | OCT-based endometrial detection and analysis method, device, equipment and storage medium |
| WO2023216594A1 (en) * | 2022-05-09 | 2023-11-16 | 深圳迈瑞生物医疗电子股份有限公司 | Ultrasonic imaging system and method |
| EP4349244A4 (en) * | 2021-05-24 | 2024-08-21 | Tomophase Ltd | Inspection method and system directly applying noninvasive oct to endometrium, and device |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102018795A (en) * | 2010-12-24 | 2011-04-20 | 武汉大学 | Traditional Chinese medicine composition for improving endometrial receptivity |
| WO2011149985A1 (en) * | 2010-05-24 | 2011-12-01 | Nanovalent Pharmaceuticals, Inc. | Polymerized shell lipid microbubbles and uses thereof |
| CN202821430U (en) * | 2012-10-10 | 2013-03-27 | 陈智毅 | Device used for forecasting pregnancy outcomes based on three dimensional-Ultrasound (3D-US) |
| CN108364293A (en) * | 2018-04-10 | 2018-08-03 | 复旦大学附属肿瘤医院 | A kind of on-line training thyroid tumors Ultrasound Image Recognition Method and its device |
| CN108986073A (en) * | 2018-06-04 | 2018-12-11 | 东南大学 | A kind of CT image pulmonary nodule detection method based on improved Faster R-CNN frame |
| CN109410194A (en) * | 2018-10-19 | 2019-03-01 | 山东省计算中心（国家超级计算济南中心） | A kind of cancer of the esophagus pathology image processing method based on deep learning |
-
2019
- 2019-03-22 CN CN201910223197.8A patent/CN109805963B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2011149985A1 (en) * | 2010-05-24 | 2011-12-01 | Nanovalent Pharmaceuticals, Inc. | Polymerized shell lipid microbubbles and uses thereof |
| CN102018795A (en) * | 2010-12-24 | 2011-04-20 | 武汉大学 | Traditional Chinese medicine composition for improving endometrial receptivity |
| CN202821430U (en) * | 2012-10-10 | 2013-03-27 | 陈智毅 | Device used for forecasting pregnancy outcomes based on three dimensional-Ultrasound (3D-US) |
| CN108364293A (en) * | 2018-04-10 | 2018-08-03 | 复旦大学附属肿瘤医院 | A kind of on-line training thyroid tumors Ultrasound Image Recognition Method and its device |
| CN108986073A (en) * | 2018-06-04 | 2018-12-11 | 东南大学 | A kind of CT image pulmonary nodule detection method based on improved Faster R-CNN frame |
| CN109410194A (en) * | 2018-10-19 | 2019-03-01 | 山东省计算中心（国家超级计算济南中心） | A kind of cancer of the esophagus pathology image processing method based on deep learning |
Non-Patent Citations (2)
| Title |
|---|
| M. S. NEOFYTOU等: "COLOR BASED TEXTURE - CLASSIFICATION OF HYSTEROSCOPY IMAGES OF THE ENDOMETRIUM", 《IEEE》 * |
| 武梅等: "子宫内膜异位症中医证型研究进展", 《中华中医药杂志》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112215843A (en) * | 2019-12-31 | 2021-01-12 | 无锡祥生医疗科技股份有限公司 | Ultrasonic intelligent imaging navigation method and device, ultrasonic equipment and storage medium |
| CN112215843B (en) * | 2019-12-31 | 2021-06-11 | 无锡祥生医疗科技股份有限公司 | Ultrasonic intelligent imaging navigation method and device, ultrasonic equipment and storage medium |
| CN111436972A (en) * | 2020-04-13 | 2020-07-24 | 王时灿 | Three-dimensional ultrasonic gynecological disease diagnosis device |
| CN111768379A (en) * | 2020-06-29 | 2020-10-13 | 深圳度影医疗科技有限公司 | Standard section detection method of three-dimensional uterine ultrasound image |
| CN112949723A (en) * | 2021-03-08 | 2021-06-11 | 西安美佳家医疗科技有限责任公司 | Endometrium pathology image classification method |
| CN112949723B (en) * | 2021-03-08 | 2023-02-14 | 西安交通大学医学院第一附属医院 | Endometrium pathology image classification method |
| EP4349244A4 (en) * | 2021-05-24 | 2024-08-21 | Tomophase Ltd | Inspection method and system directly applying noninvasive oct to endometrium, and device |
| CN113520317A (en) * | 2021-07-05 | 2021-10-22 | 汤姆飞思（香港）有限公司 | OCT-based endometrial detection and analysis method, device, equipment and storage medium |
| WO2023216594A1 (en) * | 2022-05-09 | 2023-11-16 | 深圳迈瑞生物医疗电子股份有限公司 | Ultrasonic imaging system and method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109805963B (en) | 2022-07-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109805963A (en) | The determination method and system of one Endometrium parting | |
| CN106022237B (en) | A kind of pedestrian detection method of convolutional neural networks end to end | |
| CN108805070A (en) | A kind of deep learning pedestrian detection method based on built-in terminal | |
| CN109886359A (en) | Small target detecting method and detection model based on convolutional neural networks | |
| CN107818302A (en) | Non-rigid multi-scale object detection method based on convolutional neural network | |
| CN110097559A (en) | Eye fundus image focal area mask method based on deep learning | |
| CN107105310A (en) | Figure image replacement method, device and a kind of recording and broadcasting system in a kind of net cast | |
| CN109035779A (en) | Freeway traffic flow prediction technique based on DenseNet | |
| CN113420643B (en) | Lightweight underwater target detection method based on depth separable cavity convolution | |
| CN110555836A (en) | Automatic identification method and system for standard fetal section in ultrasonic image | |
| CN109886928A (en) | A kind of target cell labeling method, device, storage medium and terminal device | |
| CN105210085A (en) | Image labeling using geodesic features | |
| CN109636846B (en) | Target positioning method based on cyclic attention convolution neural network | |
| CN110458082A (en) | A kind of city management case classification recognition methods | |
| CN110415230A (en) | A kind of CT sectioning image semantic segmentation system and method based on deep learning | |
| CN109753853A (en) | One kind being completed at the same time pedestrian detection and pedestrian knows method for distinguishing again | |
| CN115205667A (en) | Dense target detection method based on YOLOv5s | |
| CN110648309B (en) | Method and related equipment for generating anti-network synthesized erythrocyte image based on condition | |
| CN113537173B (en) | Face image authenticity identification method based on face patch mapping | |
| CN116188703B (en) | Building engineering visual management system based on BIM | |
| CN108053407A (en) | Data processing method and data handling system | |
| JP2008198078A (en) | Image detection device | |
| CN109102018B (en) | The multi-tag classification method of Gynecological secretion image | |
| CN109241893A (en) | Road selection method, device and readable storage medium storing program for executing based on artificial intelligence technology | |
| CN113052217A (en) | Prediction result identification and model training method and device thereof, and computer storage medium |
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 |