CN109805963B - Method and system for judging endometrium typing - Google Patents
Method and system for judging endometrium typing Download PDFInfo
- Publication number
- CN109805963B CN109805963B CN201910223197.8A CN201910223197A CN109805963B CN 109805963 B CN109805963 B CN 109805963B CN 201910223197 A CN201910223197 A CN 201910223197A CN 109805963 B CN109805963 B CN 109805963B
- Authority
- CN
- China
- Prior art keywords
- endometrium
- image
- typing
- uterus
- longitudinal section
- 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.)
- Active
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a method and a system for judging endometrium typing, wherein the method comprises the following steps: acquiring a standard longitudinal section image of the uterus; and analyzing the standard longitudinal section image of the uterus by utilizing an endometrium typing model to obtain an endometrium typing result of the standard longitudinal section image of the uterus, wherein the endometrium typing model is generated by training of the standard longitudinal section image of the uterus with the labeled endometrium type. In the invention, the endometrium typing result can be automatically obtained by analyzing the standard longitudinal section image of the uterus by utilizing the pre-established endometrium typing model, so that a doctor does not need to perform manual analysis, and the precision and the efficiency of endometrium typing judgment are improved.
Description
Technical Field
The invention relates to the technical field of information processing, in particular to a method and a system for endometrium typing.
Background
Endometrial typing is important for the success of IVF (in vitro fertilization) and the delivery of gametes within the oviduct. If the endometrium is not type a on or the day before the day of aspiration of the egg in the IVF cycle, implantation of the embryo will not occur or the implantation rate will be low.
At present, the analysis and diagnosis of endometrium is mainly that a doctor judges the endometrium type according to the obtained standard uterine longitudinal section and the existing knowledge. Therefore, in the prior art, the judgment precision is easy to reduce mainly depending on the experience of doctors, and the working efficiency of manual judgment is low.
Disclosure of Invention
In order to solve the problems, the invention provides a method and a system for judging endometrium typing, which realize the improvement of precision and efficiency of endometrium typing judgment.
In order to achieve the purpose, the invention provides the following technical scheme:
a method for determining endometrial typing, the method comprising:
acquiring a standard longitudinal section image of the uterus;
and analyzing the standard longitudinal section image of the uterus by utilizing an endometrium typing model to obtain an endometrium typing result of the standard longitudinal section image of the uterus, wherein the endometrium typing model is generated by training of the standard longitudinal section image of the uterus with the labeled endometrium type.
Optionally, the acquiring of the uterus standard longitudinal section image comprises:
acquiring an ultrasound scanning image of a uterus;
and determining a standard longitudinal section image of the uterus according to the image characteristics of the three layers of the uterus in the ultrasonic scanning image.
Optionally, the method further comprises:
and creating an endometrium typing model based on the standard longitudinal section image of the uterus marked with the endometrium type.
Optionally, the creating an endometrium typing model based on the standard longitudinal section image of the uterus with the labeled endometrium type comprises:
determining the standard longitudinal section image of the uterus with the marked endometrium type as a training sample;
setting a neural network model, carrying out detection and classification training on the neural network model based on the training samples, and adjusting the neural network model based on the training results to obtain an endometrium typing model, wherein the neural network model is constructed by utilizing a DenseBlock structure in a DenseNet network.
Optionally, the analyzing the standard uterine longitudinal section by using an endometrial typing model to obtain an endometrial typing result of the standard uterine longitudinal section image, including:
detecting the standard longitudinal section of the uterus by using the endometrium typing model to obtain an endometrium area;
and extracting image features of the endometrium area, and analyzing the extracted image features through the endometrium typing model to obtain an endometrium typing result of the standard longitudinal section image of the uterus.
Optionally, the setting a neural network model, performing detection and classification training on the neural network model based on the training samples, and adjusting the neural network model based on a detection training result to obtain an endometrium typing model includes:
constructing a multitask neural network for endometrium typing;
inputting the training sample into the neural network, performing convolution layer and pooling layer operation on the training sample to expand the number of image channels, acquiring deep image features of different sizes, and respectively inputting the deep image features into a detection branch and a classification branch of the neural network;
obtaining a detection frame through training of the detection branch, wherein the detection frame is used for detecting an image area aiming at endometrium;
and classifying the image region obtained by the detection frame through training of the classification branch to obtain the type of the endometrium.
A system for determining endometrial typing, the method comprising:
the acquisition unit is used for acquiring a standard longitudinal section image of the uterus;
and the analysis unit is used for analyzing the standard longitudinal section image of the uterus by utilizing an endometrium typing model to obtain an endometrium typing result of the standard longitudinal section image of the uterus, wherein the endometrium typing model is generated by training of the standard longitudinal section image of the uterus with the labeled endometrium type.
Optionally, the obtaining unit includes:
an image acquisition subunit for acquiring an ultrasound scanning image for the uterus;
the image determining subunit is used for determining a standard longitudinal section image of the uterus according to the image characteristics of the three layers of the uterus in the ultrasonic scanning image;
wherein, this system still includes:
the creating unit is used for creating an endometrium typing model based on the standard longitudinal section image of the uterus marked with the endometrium type;
wherein the creating unit includes:
the determining subunit is used for determining the standard longitudinal section image of the uterus marked with the endometrium type as a training sample;
and the training subunit is used for setting a neural network model, detecting and classifying the neural network model based on the training samples, adjusting the neural network model based on the training results, and obtaining an endometrium typing model, wherein the neural network model is constructed by utilizing a Dense Block structure in a DenseNet network.
Optionally, the analysis unit comprises:
the detection subunit is used for detecting the standard longitudinal section of the uterus by using the endometrium typing model to obtain an endometrium area;
and the typing subunit is used for extracting image characteristics of the endometrial area, analyzing the extracted image characteristics through the endometrial typing model and obtaining an endometrial typing result of the standard uterine longitudinal section image.
Optionally, the training subunit comprises:
the network building subunit is used for building a multitask neural network for endometrium typing;
the image processing subunit is used for inputting the training sample into the neural network, performing convolution layer and pooling layer operation on the training sample to expand the number of image channels, acquiring deep image features with different sizes, and respectively inputting the deep image features into a detection branch and a classification branch of the neural network;
a detection training subunit, configured to obtain a detection frame through training of the detection branch, where the detection frame is used to detect an image region for an endometrium;
and the classification training subunit is used for classifying the image area obtained by the detection frame through training of the classification branch to obtain the type of the endometrium.
Compared with the prior art, the method and the system for judging the endometrium typing provided by the invention have the advantages that after the standard longitudinal section image of the uterus is obtained, the standard longitudinal section image of the uterus is analyzed by utilizing the pre-established endometrium typing model, and the endometrium typing result can be automatically obtained, so that a doctor does not need to perform manual analysis, and the precision and the efficiency of judging the endometrium typing are improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
FIG. 1 is a schematic flow chart of a method for determining endometrial typing according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a network structure for endometrium typing according to an embodiment of the present invention;
fig. 3 is a schematic structural diagram of a system for determining endometrial typing according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The terms "first" and "second," and the like in the description and claims of the present invention and the above-described drawings are used for distinguishing between different objects and not for describing a particular order. Furthermore, the terms "comprising" and "having," as well as any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not set forth for a listed step or element but may include steps or elements not listed.
In an embodiment of the present invention, a method for determining endometrial typing is provided, and referring to fig. 1, the method includes:
s101, acquiring a standard longitudinal section image of the uterus.
When acquiring the standard longitudinal section image of the uterus, the method can be realized according to the following steps:
s1011, acquiring an ultrasonic scanning image for the uterus;
and S1012, determining a standard longitudinal section image of the uterus according to the image characteristics of the three layers of the uterus in the ultrasonic scanning image.
The wall of uterus is divided into three layers, and the outer layer is serous membrane, namely the endometrium; the middle layer is a strong and thick muscle layer, namely an myometrium, and consists of smooth muscles; the inner layer is the mucosa, i.e. the endometrium. The uterine cavity line is displayed in the middle of the ultrasonic scanning image of the uterus, the uterine cavity line is an endometrial line, and the upper side and the lower side of an endometrium are divided into an myometrium. And judging the standard longitudinal section of the uterus according to the image characteristics of the three layers of the uterus, and acquiring the standard longitudinal section of the uterus from a series of images scanned in real time.
S102, analyzing the uterus standard longitudinal section image by using an endometrium typing model to obtain an endometrium typing result of the uterus standard longitudinal section image.
Wherein, the endometrium typing mode is generated by the standard longitudinal section image training of the uterus marked with the endometrium type. In the embodiment of the invention, after the standard longitudinal section image of the uterus is obtained, the standard longitudinal section image of the uterus is not subjected to typing judgment through the experience of doctors in the prior art, but is input into a pre-created endometrium typing model as input information, and the model is used for automatically extracting and analyzing the characteristics of the image to obtain the endometrium typing result.
In the typing determination using the endometrium typing model, the process may include the steps of:
s1021, detecting the standard longitudinal section of the uterus by using an endometrium typing model to obtain an endometrium area;
s1022, extracting image characteristics of the endometrium area, and analyzing the extracted image characteristics through the endometrium typing model to obtain an endometrium typing result of the standard uterine longitudinal section image.
And inputting the standard longitudinal section image of the uterus into the endometrium typing model, wherein the model can detect the area where the endometrium is located, and the type of the endometrium in the image is given according to the image characteristics shown in the area. The image features include, but are not limited to, gray information, edge information, and the like of an image, and generally, the features need to be further abstracted and combined to obtain more accurate type judgment.
The classification of endometrium includes type A, type B and type C. Type A is represented as a multilayer three-line type endometrium, comprising a strong echogenic endometrium outer layer, a middle line of a uterine cavity and a low echogenic endometrium between the two; type B is represented by the middle intima of moderate echo, the outer intima of moderate echo, and the midline of the uterine cavity where the echo is low; type C endometrium, all layers (including endometrium, ectointimal layer, and midline of the uterine cavity) all appear as moderate echogenicity, unclear midline of the uterine cavity.
In an embodiment of the present invention, the automatic endometrium typing algorithm automatically identifies the type of endometrium by using a deep learning network, and therefore, in another embodiment of the present invention, the method further comprises:
s201, creating an endometrium typing model based on the standard longitudinal section image of the uterus marked with the endometrium type.
In creating the endometrial typing model, the process may include the steps of:
s2011, determining the standard longitudinal section image of the uterus with the marked endometrium type as a training sample;
s2022, setting a neural network model, carrying out detection and classification training on the neural network model based on the training samples, and adjusting the neural network model based on the detection training result to obtain an endometrium typing model.
In the embodiment of the invention, the deep learning neural network technology is adopted for creating the endometrium typing model, and in order to improve the precision of typing judgment, a deep learning network model suitable for automatic endometrium typing, namely the endometrium typing model, is constructed by preferably utilizing a Dense Block structure in a DenseNet network.
The Dense Block directly connects all layers on the premise of ensuring maximum information transmission between the layers in the network, and the structure lightens the phenomenon that the gradient disappears in the deep learning neural network training process, strengthens the transfer of image characteristics, more effectively multiplexes low-layer image characteristics, and is suitable for processing the uterine standard longitudinal section images.
In a conventional convolutional neural network, if the network has L layers, there are L connections, but in a DenseNet network there are L (L +1)/2 connections, with the input of each layer coming from the output of all previous layers. Thanks to the design of the Dense Block, the number of feature images output by each convolutional layer is small, the number of DenseNet parameters can be reduced, the transmission of features and gradients is more effective by the connection mode, and the training of the network is more effective.
Specifically, the setting of the neural network model, the detection and classification training of the neural network model based on the training samples, and the adjustment of the neural network model based on the detection training results to obtain the endometrium typing model includes:
s301, building a multitask neural network for endometrium typing;
s302, inputting the training sample into the neural network, performing convolution layer and pooling layer operation on the training sample to expand the number of image channels, obtaining deep image features of different sizes, and respectively inputting the deep image features into a detection branch and a classification branch of the neural network;
s303, obtaining a detection frame through training of the detection branch, wherein the detection frame is used for detecting an image area aiming at endometrium;
s304, classifying the image area obtained by the detection frame through training of the classification branch to obtain the type of the endometrium.
And training the detection branch and the classification branch to enable the trained neural network to extract image features suitable for classifying endometrium in the image so as to obtain the type of endometrium.
The endometrium-typed multitask neural network constructed in the above step S301 adopts a Dense Block structure in a DenseNet network.
The training sample is a standard longitudinal section image of the uterus with the labeled endometrium type, namely the training sample is used as data required for network training, and the endometrium type in the image can be embodied in the image in a label form. I.e. the image form of the standard longitudinal section image of the uterus with the marked type needs to be unified. According to the position, the shape and the characteristics of the endometrium in the standard longitudinal section image of the uterus, the area where the endometrium is located can be marked by a rectangular frame, coordinate values of the upper left corner and the lower right corner of the rectangle are recorded, and then the type of the endometrium is marked, for example: the type a intima label is 1, the type B intima label is 2, and the type C intima label is 3, so that one standard longitudinal section image of uterus has only one definite class label.
The endometrium typing network structure in the technical scheme provided by the invention is shown in fig. 2, and comprises two branches for detection and classification, wherein a common network part of the two branches is a Dense Block and a pooling module which are included in a dotted line frame in fig. 2, and is called backsbone for short, namely, relevant image characteristics of an input image after Dense Block and pooling are utilized in the training process of the detection branch and the classification branch. Specifically, the method comprises the following steps:
in fig. 2, a training sample, i.e., a standard longitudinal sectional image of uterus, is input, and the number of image channels is expanded by one convolutional layer; then extracting depth image features of different sizes in the image by using Dense Block and pooling, wherein pooling is down-sampling and is used for reducing the size of the image, and the pooling represents a 2 x 2 maximum pooling layer with the step length of 2, namely a 2 x 2 sliding window is adopted to slide on the image, and the step length is the sliding pixel length of the window each time; multiplexing the characteristics of the images of the lower layer in the backbone in the detection output branch, up-sampling the characteristics for gradually enlarging the size of the characteristic diagram, and finally outputting a rectangular frame for detecting the endometrium, wherein the size of the output image is the same as that of the input image; and (3) further extracting and abstractly combining image features by a Dense Block in the type output branch, obtaining probability values of the input images belonging to different types of intima through global mean pooling, and finally outputting the intima types according to the probability values, namely taking the corresponding type with the maximum probability value as an output result.
The network simultaneously trains and detects and classifies two subtasks, the tasks are mutually supervised and complemented, the network can more easily and correctly extract the image characteristics of the endometrial region, and the detection and classification precision is improved. During training, detection and classification errors are propagated reversely at the same time, updating of the weight of the convolution layer in the backbone is supervised, a neural network needs a large amount of data to carry out repeated iterative training, the trained network can automatically extract image features which are most suitable for intima classification in images, and predicted endometrial positions and typing results are obtained through analysis and calculation of the features.
It should be noted that in fig. 2, a word pattern of x 4 is marked below the sense Block and the pooling module and below the sense Block and the upsampling module in the detection branch, which characterizes that the above two processes can be performed with multiple iterative processes, and in fig. 2, 4 iterative training processes are performed, so that the training result is more accurate, and the iteration number can also be determined according to a specific situation, for example, the specific iteration number is determined according to a loss function.
Therefore, the invention adopts the multitask network to monitor the position of the inner membrane and judge the type of the inner membrane, the multitask network can obtain better characteristics under limited data, a plurality of tasks are mutually supplemented, the hidden characteristics in the image are easy to find, more tasks introduce richer supervision information from different aspects, thereby improving the algorithm precision and obtaining more credible classification results.
Because the input training sample images contain abundant information, the deep learning network is utilized to automatically extract image features with small intra-class difference and large inter-class difference from the training sample images, and then correct classification decisions are made. The loss function is an important link for guiding the network training direction, and the loss function commonly used for the classification problem has minimum mean square error, cross entropy and the like. In the single label classification problem in the embodiment of the present invention, cross entropy may be selected as a network loss function. To improve the nonlinear expression capability of the model, an activation function may be added after each convolutional layer. In the deep learning neural network training process, the number of data sets is an important influence factor for the expected effect and robustness of network training output, and due to the fact that the data volume of ultrasonic images is limited, methods such as translation, cutting and turning need to be adopted in the training process to enhance the number, and the number of samples is increased. Since only the endometrium typing result needs to be obtained, the detected position of the endometrium is not output, and only the classification result is output.
The invention provides a method for judging endometrium typing, which is characterized in that after a standard longitudinal section image of an uterus is obtained, the standard longitudinal section image of the uterus is analyzed by utilizing a pre-established endometrium typing model, and an endometrium typing result can be automatically obtained, so that a doctor does not need to perform manual analysis, and the precision and the efficiency of endometrium typing judgment are improved.
In another embodiment of the present invention, there is provided a system for determining endometrial typing, referring to fig. 3, the system comprising:
the acquisition unit 10 is used for acquiring a standard longitudinal section image of the uterus;
and the analysis unit 20 is configured to analyze the standard uterine longitudinal section image by using an endometrial typing model, and obtain an endometrial typing result of the standard uterine longitudinal section image, wherein the endometrial typing model is generated by training of the standard uterine longitudinal section image labeled with an endometrial type.
The invention provides a judging system for endometrium typing, wherein an acquisition unit is used for analyzing a standard longitudinal section image of an uterus by utilizing a pre-established endometrium typing model in an analysis unit after the standard longitudinal section image of the uterus is obtained, and an endometrium typing result can be automatically obtained, so that a doctor is not required to perform manual analysis, and the precision and the efficiency of endometrium typing judgment are improved.
On the basis of the above embodiment, the acquisition unit 10 includes:
an image acquisition subunit for acquiring an ultrasound scanning image for the uterus;
and the image determining subunit is used for determining a standard longitudinal section image of the uterus according to the image characteristics of the three layers of the uterus in the ultrasonic scanning image.
On the basis of the above embodiment, the system for determining endometrial typing further comprises:
and the creating unit is used for creating an endometrium typing model based on the standard longitudinal section image of the uterus marked with the endometrium type.
On the basis of the above embodiment, the creating unit includes:
the determining subunit is used for determining the standard longitudinal section image of the uterus with the marked endometrium type as a training sample;
and the training subunit is used for setting a neural network model, carrying out detection and classification training on the neural network model based on the training samples, and adjusting the neural network model based on the training result to obtain an endometrium typing model, wherein the neural network model is constructed by utilizing a Dense Block structure in a DenseNet network.
On the basis of the above embodiment, the analysis unit 20 in the system includes:
the detection subunit is used for detecting the standard longitudinal section of the uterus by using the endometrium typing model to obtain an endometrium area;
and the typing subunit is used for extracting image characteristics of the endometrium area, analyzing the extracted image characteristics through the endometrium typing model and obtaining an endometrium typing result of the standard longitudinal section image of the uterus.
In the above embodiment, the training subunit in the creating unit includes:
the network building subunit is used for building a multitask neural network for endometrium typing;
the image processing subunit is used for inputting the training sample into the neural network, performing convolution layer and pooling layer operation on the training sample to expand the number of image channels, acquiring deep image features with different sizes, and respectively inputting the deep image features into a detection branch and a classification branch of the neural network;
a detection training subunit, configured to obtain a detection frame through training of the detection branch, where the detection frame is used to detect an image region for an endometrium;
and the classification training subunit is used for classifying the image area obtained by the detection frame through training of the classification branch to obtain the type of the endometrium.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A method for determining endometrial typing, comprising:
acquiring a standard longitudinal section image of the uterus;
analyzing the standard longitudinal section image of the uterus by utilizing an endometrium typing model to obtain an endometrium typing result of the standard longitudinal section image of the uterus, wherein the endometrium typing model is generated by training of the standard longitudinal section image of the uterus with the labeled endometrium type;
the endometrium typing model comprises a detection branch and a classification branch, the detection branch and the classification branch are provided with a public network part backbone, the public network part backbone comprises a Dense Block and a pooling module, and low-level image characteristics in the backbone are multiplexed in the detection branch;
the typing results comprise types A, B and C; the A type is a multilayer three-line type endometrium and comprises an endometrium outer layer with strong echo, a uterine cavity midline and a low-echo endometrium between the endometrium outer layer and the uterine cavity midline; the type B is a middle intima layer with medium echo, an outer intima layer with medium echo, and the middle line of the uterine cavity with low echo; the type C is that the endometrium, the outer endometrium and the middle uterine cavity line are all moderate echoes, and the middle uterine cavity line is unclear.
2. The method of claim 1, wherein said obtaining a standard longitudinal sectional image of the uterus comprises:
acquiring an ultrasound scan image for the uterus;
and determining a standard longitudinal section image of the uterus according to the image characteristics of the three layers of the uterus in the ultrasonic scanning image.
3. The method of claim 1, further comprising:
and creating an endometrium typing model based on the standard longitudinal section image of the uterus marked with the endometrium type.
4. The method of claim 3, wherein creating an endometrium typing model based on the standard longitudinal sectional image of the uterus for the labeled endometrium type comprises:
determining the standard longitudinal section image of the uterus with the marked endometrium type as a training sample;
setting a neural network model, carrying out detection and classification training on the neural network model based on the training samples, and adjusting the neural network model based on the training results to obtain an endometrium typing model, wherein the neural network model is constructed by utilizing a Dense Block structure in a DenseNet network.
5. The method of claim 4, wherein said analyzing said standard uterine longitudinal section using an endometrium typing model to obtain endometrium typing results of said standard uterine longitudinal section image comprises:
detecting the standard longitudinal section of the uterus by using the endometrium typing model to obtain an endometrium area;
and extracting image features of the endometrium area, and analyzing the extracted image features through the endometrium typing model to obtain an endometrium typing result of the standard longitudinal section image of the uterus.
6. The method according to claim 5, wherein the setting a neural network model, performing detection and classification training on the neural network model based on the training samples, and adjusting the neural network model based on the detection training result to obtain an endometrium typing model comprises:
constructing a multitask neural network for endometrium typing;
inputting the training sample into the neural network, performing convolution layer and pooling layer operation on the training sample to expand the number of image channels, acquiring deep image features of different sizes, and respectively inputting the deep image features into a detection branch and a classification branch of the neural network;
obtaining a detection frame through training of the detection branch, wherein the detection frame is used for detecting an image area aiming at endometrium;
and through training the classification branch, classifying the image area obtained by the detection frame to obtain the type of the endometrium.
7. An endometrial typing assessment system, comprising:
the acquisition unit is used for acquiring a standard longitudinal section image of the uterus;
the analysis unit is used for analyzing the standard longitudinal section image of the uterus by utilizing an endometrium typing model to obtain an endometrium typing result of the standard longitudinal section image of the uterus, wherein the endometrium typing model is generated by training of the standard longitudinal section image of the uterus with the labeled endometrium type;
the endometrium typing model comprises a detection branch and a classification branch, the detection branch and the classification branch are provided with a public network part backbone, the public network part backbone comprises a Dense Block and a pooling module, and low-level image characteristics in the backbone are multiplexed in the detection branch;
the typing results comprise types A, B and C; the A type is a multilayer three-line type endometrium and comprises an endometrium outer layer with strong echo, a uterine cavity midline and a low-echo endometrium between the endometrium outer layer and the uterine cavity midline; the type B is a middle intima layer with medium echo, an outer intima layer with medium echo, and a midline of the uterine cavity with low echo; the type C is that the endometrium, the outer endometrium and the middle uterine cavity line are all moderate echoes, and the middle uterine cavity line is unclear.
8. The system of claim 7, wherein the obtaining unit comprises:
an image acquisition subunit for acquiring an ultrasound scanning image for the uterus;
the image determining subunit is used for determining a standard longitudinal section image of the uterus according to the image characteristics of the three layers of the uterus in the ultrasonic scanning image;
wherein, this system still includes:
the creating unit is used for creating an endometrium typing model based on the standard longitudinal section image of the uterus marked with the endometrium type;
wherein the creating unit includes:
the determining subunit is used for determining the standard longitudinal section image of the uterus with the marked endometrium type as a training sample;
and the training subunit is used for setting a neural network model, detecting and classifying the neural network model based on the training samples, adjusting the neural network model based on the training results, and obtaining an endometrium typing model, wherein the neural network model is constructed by utilizing a Dense Block structure in a DenseNet network.
9. The system of claim 8, wherein the analysis unit comprises:
the detection subunit is used for detecting the standard longitudinal section of the uterus by using the endometrium typing model to obtain an endometrium area;
and the typing subunit is used for extracting image characteristics of the endometrium area, analyzing the extracted image characteristics through the endometrium typing model and obtaining an endometrium typing result of the standard longitudinal section image of the uterus.
10. The system according to claim 9, wherein the training subunit comprises:
the network building subunit is used for building a multitask neural network for endometrium typing;
the image processing subunit is used for inputting the training sample into the neural network, performing convolution layer and pooling layer operation on the training sample to expand the number of image channels, acquiring deep image features with different sizes, and respectively inputting the deep image features into a detection branch and a classification branch of the neural network;
a detection training subunit, configured to obtain a detection frame through training of the detection branch, where the detection frame is used to detect an image region for an endometrium;
and the classification training subunit is used for classifying the image area obtained by the detection frame through training of the classification branch to obtain the type of the endometrium.
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 CN109805963A (en) | 2019-05-28 |
| CN109805963B true 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) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112288742B (en) * | 2019-12-31 | 2021-11-19 | 无锡祥生医疗科技股份有限公司 | Navigation method and device for ultrasonic probe, storage medium and electronic equipment |
| CN111436972A (en) * | 2020-04-13 | 2020-07-24 | 王时灿 | Three-dimensional ultrasonic gynecological disease diagnosis device |
| CN111768379B (en) * | 2020-06-29 | 2024-06-21 | 深圳度影医疗科技有限公司 | Standard section detection method for uterine three-dimensional ultrasonic image |
| 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 |
| CN113273968A (en) * | 2021-05-24 | 2021-08-20 | 汤姆飞思（香港）有限公司 | Detection method, device and system for directly applying noninvasive OCT (optical coherence tomography) to endometrium |
| 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 |
| CN117084709A (en) * | 2022-05-09 | 2023-11-21 | 深圳迈瑞生物医疗电子股份有限公司 | Ultrasonic imaging system and method |
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 |
|---|
| COLOR BASED TEXTURE - CLASSIFICATION OF HYSTEROSCOPY IMAGES OF THE ENDOMETRIUM;M. S. Neofytou等;《IEEE》;20070831;第364-367页 * |
| 子宫内膜异位症中医证型研究进展;武梅等;《中华中医药杂志》;20190131;第34卷(第1期);第253-255页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109805963A (en) | 2019-05-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109805963B (en) | Method and system for judging endometrium typing | |
| CN110428432B (en) | Deep neural network algorithm for automatically segmenting colon gland image | |
| CN109671086A (en) | A kind of fetus head full-automatic partition method based on three-D ultrasonic | |
| CN111445481A (en) | Abdominal CT multi-organ segmentation method based on scale fusion | |
| CN111369563A (en) | Semantic segmentation method based on pyramid void convolutional network | |
| US11967181B2 (en) | Method and device for retinal image recognition, electronic equipment, and storage medium | |
| CN101081168A (en) | Method for shielding sex part on foetus image for preventing recognizing foetus sex | |
| CN111325738B (en) | Intelligent detection method and system for transverse hole peripheral cracks | |
| CN112149620A (en) | Method for constructing natural scene character region detection model based on no anchor point | |
| CN112132166A (en) | Intelligent analysis method, system and device for digital cytopathology image | |
| CN110136153A (en) | A kind of image processing method, equipment and storage medium | |
| CN111920405A (en) | Atrial fibrillation signal identification system and method | |
| CN105319223A (en) | Detection method and system of chalky rice | |
| CN114821316A (en) | Three-dimensional ground penetrating radar crack disease identification method and system | |
| CN103169506A (en) | Ultrasonic diagnosis device and method capable of recognizing liver cancer automatically | |
| CN116740528A (en) | Shadow feature-based side-scan sonar image target detection method and system | |
| CN111754485A (en) | Artificial intelligence ultrasonic auxiliary system for liver | |
| CN113469961A (en) | Neural network-based carpal tunnel image segmentation method and system | |
| CN115471745A (en) | Network model and device for plant identification and electronic equipment | |
| CN116416221A (en) | Ultrasonic image analysis method | |
| CN113643263A (en) | Identification method and system for upper limb bone positioning and forearm bone fusion deformity | |
| CN116322521A (en) | Ultrasonic imaging method and ultrasonic imaging system for midnight pregnancy fetus | |
| CN116823829B (en) | Medical image analysis method, medical image analysis device, computer equipment and storage medium | |
| WO2020037563A1 (en) | Method for ultrasound imaging and related equipment | |
| CN118229540B (en) | Ultrasonic image intelligent processing method of head-mounted display equipment |
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 |