CN110473208A - Structural form method of discrimination, computer equipment and storage medium - Google Patents
Structural form method of discrimination, computer equipment and storage medium Download PDFInfo
- Publication number
- CN110473208A CN110473208A CN201910701918.1A CN201910701918A CN110473208A CN 110473208 A CN110473208 A CN 110473208A CN 201910701918 A CN201910701918 A CN 201910701918A CN 110473208 A CN110473208 A CN 110473208A
- Authority
- CN
- China
- Prior art keywords
- network
- segmentation
- object construction
- segmentation result
- heart
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10116—X-ray image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—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
- G06T2207/30048—Heart; Cardiac
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Image Processing (AREA)
Abstract
This application involves a kind of structural form method of discrimination, computer equipment and storage mediums, since computer equipment is that the segmentation result of object construction is carried out to the judgement of structural form using trained Morphological Identification network model in advance, and the Morphological Identification network is for judging each structural form information of human body, in this way, the shape information that the object construction is judged using the differentiation network of special training, can greatly improve the precision and efficiency of the differentiation of object construction form.
Description
Technical field
This application involves field of medical technology, more particularly to a kind of structural form method of discrimination, computer equipment and deposit
Storage media.
Background technique
Heart is a very important organ in human body, due to the shape of each human heart of the difference of physiological status
State can also have certain difference.Common physiological cardiac form has following several: dropping heart, horizocardia, pear-shaped heart, boot-shaped
The heart, Pu great Xin, the flask heart, spherical center etc..
The accurate judgement of cardiac shape, which assists in diagnosis, whether there is the heart disease of early stage, or, if
In the presence of the obesity etc. that may oppress heart, so that the accurate judgement of cardiac shape is to the diagnosis of cardiac-related diseases with great
Meaning.Currently, the method for cardiac shape judgement is usually to pass through x-ray chest radiograph to navigate to heart region, then artificial judgement
The form in the region is so that it is determined that the form of heart out.
But there is a problem of that precision and efficiency are lower to the differentiation of cardiac shape in the prior art.
Summary of the invention
Based on this, it is necessary to which to the differentiation of cardiac shape, there are precision and the lower skill of efficiency for above-mentioned in the prior art
Art problem provides a kind of structural form method of discrimination, computer equipment and storage medium.
In a first aspect, the embodiment of the present application provides a kind of structural form method of discrimination, this method comprises:
Obtain the segmentation result of object construction to be discriminated；
The segmentation result of object construction to be discriminated is input in Morphological Identification network, the form letter of object construction is obtained
Breath；Wherein, Morphological Identification network is used to judge the shape information of each structure of human body.
Second aspect, the embodiment of the present application provide a kind of structural form discriminating gear, which includes:
Structure obtains module, for obtaining the segmentation result of object construction to be discriminated；
Morphological Identification module is obtained for the segmentation result of object construction to be discriminated to be input in Morphological Identification network
The shape information of object construction；Wherein, Morphological Identification network is used to judge the shape information of each structure of human body.
The third aspect, the embodiment of the present application provide a kind of computer equipment, including memory and processor, which deposits
Computer program is contained, which realizes the either step of the above first aspect embodiment when executing computer program.
Fourth aspect, the embodiment of the present application provide a kind of computer readable storage medium, are stored thereon with computer program,
The computer program realizes the either step of the above first aspect embodiment when being executed by processor.
A kind of structural form method of discrimination, computer equipment and storage medium provided by the present application, due to computer equipment
It is the segmentation result of object construction to be carried out to the judgement of structural form using trained Morphological Identification network model in advance, and be somebody's turn to do
Morphological Identification network is for judging each structural form information of human body, in this way, the differentiation network using special training judges
The shape information of the object construction can greatly improve the precision and efficiency of the differentiation of object construction form.
Detailed description of the invention
Fig. 1 is the applied environment figure for the structural form method of discrimination that one embodiment provides；
Fig. 2 is the flow diagram for the structural form method of discrimination that one embodiment provides；
Fig. 3 is the flow diagram for the structural form method of discrimination that one embodiment provides；
Fig. 4 is the flow diagram for the structural form method of discrimination that one embodiment provides；
Fig. 5 is the flow diagram for the structural form method of discrimination that one embodiment provides；
Fig. 6 is the flow diagram for the structural form method of discrimination that one embodiment provides；
Fig. 7 is the flow diagram for the structural form method of discrimination that one embodiment provides；
Fig. 8 is the structural form method of discrimination complete diagram that one embodiment provides；
Fig. 8 a is the structural form method of discrimination instance graph that one embodiment provides；
Fig. 9 is the structural block diagram for the structural form discriminating gear that one embodiment provides；
Figure 10 is the structural block diagram for the structural form discriminating gear that one embodiment provides；
Figure 11 is the structural block diagram for the structural form discriminating gear that one embodiment provides；
Figure 12 is the structural block diagram for the structural form discriminating gear that one embodiment provides；
Figure 13 is the structural block diagram for the structural form discriminating gear that one embodiment provides.
Specific embodiment
It is with reference to the accompanying drawings and embodiments, right in order to which the objects, technical solutions and advantages of the application are more clearly understood
The application is further elaborated.It should be appreciated that specific embodiment described herein is only used to explain the application, not
For limiting the application.
A kind of structural form method of discrimination provided by the present application, can be applied in application environment as shown in Figure 1, this is answered
With in environment, the processor of computer equipment is for providing calculating and control ability.The memory of the computer equipment includes non-
Volatile storage medium, built-in storage.The non-volatile memory medium is stored with operating system, computer program and database.
The built-in storage provides environment for the operation of operating system and computer program in non-volatile memory medium.The computer is set
Standby database is used for the data of storage organization Morphological Identification method.The network interface of the computer equipment is used for and external its
He passes through network connection communication at equipment.To realize a kind of structural form method of discrimination when the computer program is executed by processor.
General common physiological cardiac shape dropping heart and two kinds of horizocardia, wherein dropping heart patient typically exhibits out
Physique is tall and thin, thorax is long and narrow, every the features such as position is low, and heart shadow is general less than normal and long and narrow, in position of hanging down, the heart longitudinal axis and horizontal plane
Angle is greater than 45 degree, the heart with it is small every contact surface, ambition ratio is less than 0.5 under normal conditions, and pulmonary artery section is longer, slightly dashes forward；Wherein, horizontal
The position heart is common in short and stout physique, and patient's thorax is wide and short, and diaphragm position is high, and the angle of the heart longitudinal axis and horizontal plane is small (less than 45
Degree), the contact surface of the heart and diaphragm is big, and cardiothoracic ratio is often greater than 0.5.Bulbus aortae is obvious, heart waist recess.Left anterior oblique position interventricular groove position
In diaphragmatic surface level, heart rear can be slightly Chong Die with backbone.Common pathologic heart shape pear-shaped heart, boot-shaped heart, Pu great Xin, burning
The bottle heart and spherical center, wherein pear-shaped heart indicates that pulmonary artery section protrusion and the apex of the heart upwarp, and aorta tubercle reduces or normally, shape is such as
Pyriform, being more common in right cardiac load or the chambers of the heart variation based on it, common disease has mitral lesion, atrial septal defect, lung dynamic
Arteries and veins valve is narrow, pulmonary hypertension and pulmonary heart disease etc..Wherein, boot-shaped heart refers to that border of cardiac dullness expands to left down, and waist of heart is by blunt
Angle becomes approximate right angle, makes border of cardiac dullness in boot last, narrow also seen in aorta petal because it is common in aortic incompetence
It is narrow, the also known as aortic type heart, also seen in hypertensive cardiopathy, tetralogy of Fallot.Wherein, Pu great Xin is that " heart is universal
The abbreviation of expansion ", it is a kind of performance of cardiac contour, by heart X-ray examination or heart percussion it was determined that if not having
Hydropericardium, " Pu great Xin " mean that ventricle and atrium and have apparent expansion, and when hydropericardium, heart shadow can generally increase,
But and non-cardiac increase itself, be also shown the whole heart failure caused by hypertension.Wherein, the flask heart indicates on x-ray chest radiograph, the heart
Shadow is expanded in flask sample spherical shape, and bi-ventricular increase circle is grand, and down big up small similar flask, this is the characteristic table of hydropericardium
It is existing.Wherein, spherical center refers to that chamber generally increases, and organic heart disease advanced stage (also sees newborn, pediatric cardiac X piece).Base
In the above-mentioned various heart shapes enumerated, when diagnosing heart and lung diseases, X-ray (X-Rays) rabat is due to its convenience and economy
Property be widely used at present, most physical examination and inspection can all carry out the inspection of x-ray chest radiograph, therefore for heart
The diagnosis of disease and it is found to have important meaning.At present by X-ray judge heart shape be mainly pass through estimate and
The mode of hand dipping is realized, when needing to handle a large amount of rabat data, inefficient shape inevitably occurs in the method judged manually
Condition.Therefore, the embodiment of the present application provides a kind of structural form method of discrimination, computer equipment and storage medium, it is intended to solve existing
With the presence of the lower technical problem of differentiation precision and efficiency to cardiac shape in technology.Embodiment will be passed through below and combined attached
Figure specifically carries out specifically to how the technical solution of the technical solution of the application and the application solves above-mentioned technical problem
It is bright.These specific embodiments can be combined with each other below, may be in certain realities for the same or similar concept or process
It applies in example and repeats no more.It should be noted that a kind of structural form method of discrimination provided by the present application, the execution master of Fig. 2-Fig. 8
Body is computer equipment, wherein its executing subject can also be structural form discriminating gear, and wherein the device can be by soft
The mode of part, hardware or software and hardware combining is implemented as some or all of of structural form differentiation.
To keep the purposes, technical schemes and advantages of the embodiment of the present application clearer, below in conjunction with the embodiment of the present application
In attached drawing, the technical scheme in the embodiment of the application is clearly and completely described, it is clear that described embodiment is
Some embodiments of the present application, instead of all the embodiments.
In one embodiment, Fig. 2 provides a kind of structural form method of discrimination, and what is involved is computers to set for the present embodiment
The standby detailed process that object construction shape information is determined according to the segmentation result of object construction to be discriminated.As shown in Fig. 2, the party
Method includes:
S101 obtains the segmentation result of object construction to be discriminated.
In the present embodiment, object construction to be discriminated indicates the human figure structure for needing to differentiate shape information.Example
Such as, stomach, lung, heart etc., the present embodiment does not limit this.It should be noted that the embodiment of the present application will be with heart
For be illustrated.The segmentation result that computer equipment obtains object construction to be discriminated indicates that is obtained has split
Structure, illustratively, in practical applications, the mode that computer equipment obtains the segmentation result of object construction to be discriminated can be meter
It calculates machine equipment and from database directly acquires segmented good object construction, be also possible to computer equipment in real time using setting in advance
The partitioning scheme (such as partitioning algorithm, preparatory trained segmentation network model for setting etc.) set is to object construction
It is split, can also be and directly download segmented good object construction from network, the present embodiment divides object construction
As a result acquisition modes without limitation, as long as computer equipment has got the segmentation result of object construction.
The segmentation result of object construction to be discriminated is input in Morphological Identification network by S102, obtains the shape of object construction
State information；Wherein, Morphological Identification network is used to judge the shape information of each structure of human body.
Based on the segmentation result of the object construction to be discriminated obtained in above-mentioned S101 step, computer equipment is by mesh to be discriminated
The segmentation result of mark structure is input in Morphological Identification network, obtains the shape information of object construction.Wherein, the Morphological Identification net
Network is preparatory trained, the network model of the shape information for judging each structure of human body, wherein shape information indicates target
Shape, size of structure etc. can indicate that the information of structural form, the embodiment of the present application will be carried out by shape of shape information
Explanation.
Structural form method of discrimination provided in this embodiment, since computer equipment is to adopt the segmentation result of object construction
The judgement of structural form is carried out with preparatory trained Morphological Identification network model, and the Morphological Identification network is for judging people
Each structural form information of body, it, can be in this way, judge the shape information of the object construction using the differentiation network of special training
Greatly improve the precision and efficiency of target morphology Morphological Identification.
It should be noted that Morphological Identification network can use traditional algorithm in some scenes in above-described embodiment
It carries out, such as traditional algorithm technology is using including but not limited to such as PCA, the technologies such as LDA, the present embodiment do not limit this.
Based on the above embodiment, the mode of the segmentation result of object construction, the application are obtained for above-mentioned computer equipment
Embodiment provides a kind of structural form method of discrimination, then in one embodiment, as shown in figure 3, above-mentioned S101 includes:
S201 obtains the primitive medicine image of object construction.
Computer equipment needs to get the segmentation result of object construction, then in this step, computer in the present embodiment
Equipment first obtains the primitive medicine image of object construction, wherein the primitive medicine image of object construction indicates complete to object construction
The medical image of whole acquisition, for example, if then the primitive medicine image of object construction is the X of chest when object construction is heart
Mating plate, alternatively, can also be the PET image etc. of chest, the present embodiment is not limited this, as long as computer equipment in this step
Get the medical image comprising object construction.
Primitive medicine image is input in segmentation network, obtains the segmentation result of object construction by S202；Wherein, divide
Network from medical image for being partitioned into each structure of human body.
Based on the primitive medicine image of the object construction obtained in above-mentioned S201 step, computer equipment is by the primitive medicine
Image is input in segmentation network, obtains the segmentation result of object construction, wherein the segmentation network is trained in advance, use
In the network model for each structure of human body being partitioned into from medical image, wherein input primitive medicine image in segmentation network
Before, primitive medicine image can be pre-processed, it is to further increase the robustness of object construction segmentation result, then optional
Ground, the embodiment of the present application also provides a kind of structural form method of discrimination, as shown in figure 4, in one embodiment, above-mentioned S202
Include:
S301, is normalized the primitive medicine image of object construction and/or standardization.
In the present embodiment, the primitive medicine image of object construction is normalized and/or standardization.Wherein normalizing
Change can carry out simultaneously with standardization, can also separate and one is selected to carry out.
Medical image after normalization and/or standardization is input in segmentation network, obtains object construction by S302
Segmentation result.
It, will treated that medical image is input to point after handling in above-mentioned S301 step primitive medicine image
It cuts in network, obtains the segmentation result of object construction, due to having carried out normalization or standardization to primitive medicine image,
So the input data more standard of segmentation network, can be further improved the robustness of object construction segmentation result in this way.
By taking object construction is heart as an example, it is traditionally used for the algorithm of cardiac segmentation, can be mainly divided into three kinds, one is
Threshold segmentation, this is the cardiac segmentation algorithm of the most common direct detection zone in traditional algorithm, by extracting original image
In pixel characteristic be split, it is representative such as Otsu (big saliva) Threshold Segmentation Algorithm, when which will acquire optimal threshold
It needs largely to be calculated, while being easier the influence by noise.Another is the partitioning algorithm of edge detection, should
Marginal point in algorithm detection image first generates profile by these marginal points to constitute cut zone, these are typical
Method is all to construct the differential gray scale operator such as LOG (Laplace of Gauss, Laplce Gauss) sensitive to pixel grey scale
Operator, Sobel (Suo Bai) operator, but such algorithm is all very sensitive generally for noise, and it is high to be unable to get stabilization, robustness
Segmentation result.There are also one is morphologic partitioning algorithm is based on, basic thought is the structural element degree of going with certain form
Correspondence form in amount and extraction image is to achieve the purpose that, to image analysis and identification, representative has Watersheds (to divide
Water ridge) algorithm, the algorithm problem lower there is also robustness.Therefore, through this embodiment in advance trained segmentation net
Network is partitioned into object construction from the primitive medicine image of object construction, can more efficiently obtain accurate object construction
Segmentation result.
It is carried out specifically below by training process of some embodiments to above-mentioned Morphological Identification network and segmentation network
It is bright.
As shown in figure 5, in one embodiment, the training process of above-mentioned Morphological Identification network includes:
S401 obtains the segmentation result of each structure, the shape information of each structure；The segmentation result of each structure is segmentation network
The result of output.
In the present embodiment, computer equipment will train Morphological Identification network, then computer equipment first has to obtain trained sample
This, Morphological Identification network is for determining each structural form information according to each segmentation of structures result, it is therefore desirable to be obtained big
Amount various structures segmentation result and the corresponding shape information of various structures, for example, the segmentation result of available heart and
Corresponding cardiac shape information, also available stomach segmentation result and corresponding stomach shape information, can also be lung point
Result and corresponding lung's shape information are cut, the present embodiment does not limit this, as long as the morphosis of human body.It needs
Illustrate, when obtaining training sample, the available multiple and different segmentation result of a human figure structure, and it is corresponding
Shape information, the diversity of sample can be increased in this way.Illustratively, it illustrates by heart of human body morphosis, then in reality
In the application of border, then multiple segmentation results of the available heart of computer equipment obtain the shape information of corresponding segmentation result,
Such as the shapes such as dropping heart, horizocardia, boot-shaped heart, the present embodiment to this also without limitation.Wherein, computer equipment in this step
The segmentation result of the various structures obtained is the result obtained using segmentation network.The instruction of human body various structures is used in this step
Practice sample to be trained Morphological Identification network, the Morphological Identification network obtained in application in this way can differentiate that human body is each
The shape information of structure expands Morphological Identification network application range, greatly enhances Morphological Identification network application.
S402 differentiates network according to the segmentation result of each structure and the shape information of each structure training initial configuration, obtains
Morphological Identification network.Optionally, Morphological Identification network is multilayer convolutional network.
Based on the differentiation training sample that above-mentioned S401 step obtains, computer equipment is by the segmentation result of various structures
It is input to initial configuration with the shape information of various structures to differentiate in network, according to the segmentation result of various structures and various structures
Shape information to initial configuration differentiate network be iterated training, until training, obtain above-mentioned Morphological Identification network,
Wherein the Morphological Identification network is multilayer convolutional network, such as (Visual Geometry Group-net, vision are several by vgg-net
What group network), resnet (Residual Neural Network, residual error neural network), densenet (Densely
Connected Convolutional Networks, intensive convolutional neural networks) network etc..In this way, using a large amount of various
The training sample training Morphological Identification network of structure, can make the more stable of Morphological Identification network training, more robust.
In another embodiment, as shown in fig. 6, the training process of above-mentioned segmentation network includes:
S501 obtains the medical image of each structure, the annotation results of each structure.
It is identical as above-mentioned Morphological Identification network, current embodiment require that obtaining the training sample of a large amount of segmentation networks, that is, obtain
The annotation results of the medical image and various structures of a large amount of various structures in medical image are taken, with above-mentioned Morphological Identification net
Network is identical, in order to allow the segmentation network finally obtained to divide each structure of human body, when obtaining segmentation training sample,
The mark knot of the medical image of the various structures such as available heart, lung, stomach and corresponding each structure in medical image
Fruit expands the application range of segmentation network to increase the diversity of segmentation training sample.It should be understood that the present embodiment
Obtain divide training sample when, also a variety of data of available same structure, for example, the available dropping heart of heart,
Horizocardia, the medical image of boot-shaped heart equal samples and annotation results further increase the diversity of segmentation training sample.Show
Example ground, illustrates by heart of human figure structure, and in practical applications, computer equipment obtains the various chest x-ray pieces of heart,
And the annotation results of corresponding chest x-ray piece cardiac, wherein the annotation results can be heart institute in chest x-ray on piece
It is come out in area marking, is also possible to that the region of heart shows using data, mark knot of the present embodiment to each structure
The representation of fruit is without limitation.
S502 is divided according to the annotation results of the medical image of each structure and each structure training initial segmentation network
Network.Optionally, in one embodiment, above-mentioned segmentation network is full convolutional network.
The annotation results of medical image and each structure based on each structure obtained in above-mentioned S501 step, computer are set
It is standby that the medical image of each structure and the annotation results of each structure are input in initial segmentation network, according to the medicine shadow of each structure
The annotation results of picture and each structure are iterated training to initial segmentation network, until the initial segmentation network training is good,
Obtain segmentation network.Wherein, according to the annotation results training initial segmentation network of the medical image of each structure and each structure it
Before, it needs to pre-process training sample, such as translation conversion or standardization processing etc., then in one embodiment, such as
Shown in Fig. 7, this method comprises:
The annotation results of S601, medical image and each structure to each structure carry out left and right horizontal overturning, horizontal vertical histogram
To at least one of translation transformation, transformed each image is obtained.
Wherein, the mode that the annotation results of the medical image to each structure and each structure are converted includes that left and right horizontal is turned over
Turn or horizontal vertical direction translation at least one of, i.e., the annotation results of the medical image to each structure and each structure carry out
When processing, can only carry out one of processing or two kinds of processing therein all can, the present embodiment does not limit this, this
Carried out in step transformation be equivalent to the medical image of each structure and the annotation results of each structure are converted after be formed it is new
Sample, the diversity of training sample can be increased in this way.
S602, the annotation results of medical image and each structure to each structure or transformed each image carry out normalizing
At least one of change processing, standardization processing.
The diversity that can increase sample is converted to sample in above-mentioned S601 step, but sample is continued in this step
At least one of normalization, standardization processing, can make sample data more standardize, the sample data of operating specification
Training initial segmentation network can make the faster more efficient study of network arrive the feature of sample data, wherein right in this step
The standardization processing of sample data can be to above-mentioned translation conversion after formation data be normalized, standardize at least
A kind of processing is also possible to carry out normalizing to the medical image of each structure before no processing and the annotation results of each structure
At least one of change, standardization processing, the present embodiment do not limit this.In this way, being carried out to the sample data of segmentation network flat
Transfer changes and can increase sample diversity, and standardizes to treated sample data or untreated preceding sample data
Processing can make to divide the feature that sample data is arrived in the more efficient study of network.
Illustratively, the application provides a kind of cardiac segmentation and cardiac shape differentiates complete flow chart, as shown in figure 8, tool
Body step includes:
S1: carrying out the acquisition of rabat data by X-ray, obtains initial data, and initial data includes having various cardiac shapes
Such as dropping heart, horizocardia, boot-shaped heart form.
S2: image preprocessing is carried out to initial data.
Specifically, a part of data to be chosen to be labeled as training set, the image of mark is denoted as mask (mask), and
Retain a part of rabat as test set.Training set and corresponding annotation results are done simultaneously random left and right horizontal overturning with
And translation transformation horizontal, on vertical direction, normalization and standardization then are done to image.By manually marking and
The available a large amount of training sample of data prediction, each sample include x-ray image and the corresponding mask marked.
Regard Input (input) by pretreated x-ray image, the mask by transformation is as Ground Truth (goldstandard).
S3: the building stage of cardiac segmentation model.
Above-mentioned Input is input in the model of training stage, i.e., such as above-mentioned steps 2) Input is obtained, then obtain phase
Corresponding goldstandard Ground Truth；Input is input in initial segmentation network structure, using Ground Truth as mark
Label, are split network training, and training completes that segmentation network model can be obtained.Wherein segmentation network can be deep learning
Network structure, the network used include but is not limited to Fully Convolutional Networks (full convolutional network), example
Such as: V-NET (V-shape network), U-NET (U-shaped network) and FC-Densenet (FC- intensively connects convolutional network).
S4: the building stage of cardiac shape discrimination model.
The cardiac image marked is regarded into Input, corresponding cardiac shape after standardization and normalized
Classification (Ground Truth) be input to deep neural network (use but be not limited to: VGG-Net, Resnet, Densenet) when
In, it is trained with the classification of cardiac shape as label.
S5: test phase.
It will prepare the data for being used to test by normalizing with after normalizing operation, and be input to what above-mentioned S3 step was built
The segmentation result of heart is obtained in segmentation network, the segmentation result of heart is then input to the heart that above-mentioned S4 step is built
Morphological Identification model obtains the form of heart.Wherein, in this step, in addition to using the cardiac shape discrimination model built,
When can also be using traditional technologies such as PCA, do not need to train network model when using traditional technology, it can be directly by heart
Mask carries out a series of mathematical operations to obtain the form that can judge heart in the hope of feature.
For example, obtaining three groups of examples comparative figures as shown in Figure 8 a using the step of S1-S5, such as the various realities of following table can be obtained
The shape information of the heart obtained under example.It can be in this way and automatically carry out cardiac segmentation and cardiac shape differentiation, it can be right
Common x-ray chest radiograph carries out high-precision cardiac segmentation, even the more general rabat the application of shooting quality obtain it is relatively good
Segmentation result.
| Sample | Mark label manually | Deep learning predicts cardioid | Traditional algorithm predicts cardioid |
| Example 1 | Dropping heart | Dropping heart | Dropping heart |
| Example 2 | The pears type heart | The pears type heart | The pears type heart |
| Example 3 | Horizocardia | Horizocardia | Horizocardia |
It should be understood that although each step in the flow chart of Fig. 2-8 is successively shown according to the instruction of arrow,
These steps are not that the inevitable sequence according to arrow instruction successively executes.Unless expressly stating otherwise herein, these steps
Execution there is no stringent sequences to limit, these steps can execute in other order.Moreover, at least one in Fig. 2-8
Part steps may include that perhaps these sub-steps of multiple stages or stage are not necessarily in synchronization to multiple sub-steps
Completion is executed, but can be executed at different times, the execution sequence in these sub-steps or stage is also not necessarily successively
It carries out, but can be at least part of the sub-step or stage of other steps or other steps in turn or alternately
It executes.
In one embodiment, as shown in figure 9, providing a kind of structural form discriminating gear, which includes: that structure obtains
Modulus block 10, Morphological Identification module 11, wherein
Structure obtains module 10, for obtaining the segmentation result of object construction to be discriminated；
Morphological Identification module 11 is obtained for the segmentation result of object construction to be discriminated to be input in Morphological Identification network
To the shape information of object construction；Wherein, Morphological Identification network is used to judge the shape information of each structure of human body.
A kind of structural form discriminating gear provided by the above embodiment, implementing principle and technical effect and the above method are real
It is similar to apply example, details are not described herein.
In one embodiment, as shown in Figure 10, a kind of structural form discriminating gear is provided, above structure obtains module
10, comprising: structure acquiring unit 101, cutting unit 102, wherein
Structure acquiring unit 101, for obtaining the primitive medicine image of object construction；
Cutting unit 102 obtains the segmentation knot of object construction for primitive medicine image to be input in segmentation network
Fruit；Wherein, segmentation network from medical image for being partitioned into each structure of human body.
A kind of structural form discriminating gear provided by the above embodiment, implementing principle and technical effect and the above method are real
It is similar to apply example, details are not described herein.
In one embodiment, above-mentioned cutting unit 102 is specifically used for returning the primitive medicine image of object construction
One change and/or standardization；Medical image after normalization and/or standardization is input in segmentation network, is obtained
The segmentation result of object construction.
A kind of structural form discriminating gear provided by the above embodiment, implementing principle and technical effect and the above method are real
It is similar to apply example, details are not described herein.
In one embodiment, as shown in figure 11, a kind of structural form discriminating gear, the device are provided further include:
First sample obtains module 12, for obtaining the segmentation result of each structure, the shape information of each structure；Each structure
Segmentation result is the result for dividing network output；
First training module 13, for according to the segmentation result of each structure and the shape information training initial configuration of each structure
Differentiate network, obtains Morphological Identification network.
A kind of structural form discriminating gear provided by the above embodiment, implementing principle and technical effect and the above method are real
It is similar to apply example, details are not described herein.
In one embodiment, as shown in figure 12, a kind of structural form discriminating gear, the device are provided further include:
Second sample acquisition module 14, for obtaining the medical image of each structure, the annotation results of each structure；
Second training module 15, for according to the medical image of each structure and the annotation results training initial segmentation of each structure
Network obtains segmentation network.
A kind of structural form discriminating gear provided by the above embodiment, implementing principle and technical effect and the above method are real
It is similar to apply example, details are not described herein.
In one embodiment, as shown in figure 13, a kind of structural form discriminating gear, the device are provided further include:
Conversion module 16, for the annotation results of the medical image to each structure and each structure carry out left and right horizontal overturning,
At least one of horizontal vertical direction translation transformation, obtains transformed each image；
Processing module 17, annotation results or transformed each figure for medical image and each structure to each structure
As being normalized, at least one of standardization processing.
A kind of structural form discriminating gear provided by the above embodiment, implementing principle and technical effect and the above method are real
It is similar to apply example, details are not described herein.
In one embodiment, above-mentioned segmentation network is full convolutional network.
In one embodiment, above-mentioned Morphological Identification network is multilayer convolutional network.
A kind of structural form discriminating gear provided by the above embodiment, implementing principle and technical effect and the above method are real
It is similar to apply example, details are not described herein.
Specific about structural form discriminating gear limits the limit that may refer to above for structural form method of discrimination
Fixed, details are not described herein.Modules in above structure Morphological Identification device can fully or partially through software, hardware and its
Combination is to realize.Above-mentioned each module can be embedded in the form of hardware or independently of in the processor in computer equipment, can also be with
It is stored in the memory in computer equipment in a software form, in order to which processor calls the above modules of execution corresponding
Operation.
In one embodiment, a kind of computer equipment is provided, which can be terminal, internal structure
Figure can be as shown in Figure 1 above.The computer equipment include by system bus connect processor, memory, network interface,
Display screen and input unit.Wherein, the processor of the computer equipment is for providing calculating and control ability.The computer equipment
Memory include non-volatile memory medium, built-in storage.The non-volatile memory medium is stored with operating system and calculating
Machine program.The built-in storage provides environment for the operation of operating system and computer program in non-volatile memory medium.It should
The network interface of computer equipment is used to communicate with external terminal by network connection.The computer program is executed by processor
When to realize a kind of structural form method of discrimination.The display screen of the computer equipment can be liquid crystal display or electric ink
Display screen, the input unit of the computer equipment can be the touch layer covered on display screen, be also possible to outside computer equipment
Key, trace ball or the Trackpad being arranged on shell can also be external keyboard, Trackpad or mouse etc..
It will be understood by those skilled in the art that structure shown in Fig. 1, only part relevant to application scheme is tied
The block diagram of structure does not constitute the restriction for the computer equipment being applied thereon to application scheme, specific computer equipment
It may include perhaps combining certain components or with different component layouts than more or fewer components as shown in the figure.
In one embodiment, a kind of computer equipment, including memory and processor are provided, is stored in memory
Computer program, the processor perform the steps of when executing computer program
Obtain the segmentation result of object construction to be discriminated；
The segmentation result of object construction to be discriminated is input in Morphological Identification network, the form letter of object construction is obtained
Breath；Wherein, Morphological Identification network is used to judge the shape information of each structure of human body.
A kind of computer equipment provided by the above embodiment, implementing principle and technical effect and above method embodiment class
Seemingly, details are not described herein.
In one embodiment, a kind of computer readable storage medium is provided, computer program is stored thereon with, is calculated
Machine program performs the steps of when being executed by processor
Obtain the segmentation result of object construction to be discriminated；
The segmentation result of object construction to be discriminated is input in Morphological Identification network, the form letter of object construction is obtained
Breath；Wherein, Morphological Identification network is used to judge the shape information of each structure of human body.
A kind of computer readable storage medium provided by the above embodiment, implementing principle and technical effect and the above method
Embodiment is similar, and details are not described herein.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, the computer program can be stored in a non-volatile computer
In read/write memory medium, the computer program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein,
To any reference of memory, storage, database or other media used in each embodiment provided herein,
Including non-volatile and/or volatile memory.Nonvolatile memory may include read-only memory (ROM), programming ROM
(PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM) or flash memory.Volatile memory may include
Random access memory (RAM) or external cache.By way of illustration and not limitation, RAM is available in many forms,
Such as static state RAM (SRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double data rate sdram (DDRSDRAM), enhancing
Type SDRAM (ESDRAM), synchronization link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM
(RDRAM), direct memory bus dynamic ram (DRDRAM) and memory bus dynamic ram (RDRAM) etc..
Each technical characteristic of above embodiments can be combined arbitrarily, for simplicity of description, not to above-described embodiment
In each technical characteristic it is all possible combination be all described, as long as however, the combination of these technical characteristics be not present lance
Shield all should be considered as described in this specification.
The several embodiments of the application above described embodiment only expresses, the description thereof is more specific and detailed, but simultaneously
It cannot therefore be construed as limiting the scope of the patent.It should be pointed out that coming for those of ordinary skill in the art
It says, without departing from the concept of this application, various modifications and improvements can be made, these belong to the protection of the application
Range.Therefore, the scope of protection shall be subject to the appended claims for the application patent.
Claims (10)
1. a kind of structural form method of discrimination, which is characterized in that the described method includes:
Obtain the segmentation result of object construction to be discriminated；
The segmentation result of the object construction to be discriminated is input in Morphological Identification network, the form of the object construction is obtained
Information；The Morphological Identification network is used to judge the shape information of each structure of human body.
2. the method according to claim 1, wherein the segmentation result packet for obtaining object construction to be discriminated
It includes:
Obtain the primitive medicine image of the object construction；
The primitive medicine image is input in segmentation network, the segmentation result of the object construction is obtained；The segmentation net
Network from medical image for being partitioned into each structure of human body.
3. according to the method described in claim 2, it is characterized in that, described be input to segmentation network for the primitive medicine image
In, obtain the segmentation result of the object construction, comprising:
The primitive medicine image of the object construction is normalized and/or standardization；
Medical image after normalization and/or standardization is input in the segmentation network, the object construction is obtained
Segmentation result.
4. the method according to claim 1, wherein the training process of the Morphological Identification network includes:
Obtain the segmentation result of each structure, the shape information of each structure；The segmentation result of each structure is segmentation network
The result of output；
Network is differentiated according to the segmentation result of each structure and the shape information of each structure training initial configuration, obtains institute
State Morphological Identification network.
5. according to the method described in claim 4, it is characterized in that, the training process of the segmentation network includes:
Obtain the medical image of each structure, the annotation results of each structure；
According to the annotation results of the medical image of each structure and each structure training initial segmentation network, described point is obtained
Cut network.
6. according to the method described in claim 5, it is characterized in that, the medical image according to each structure and described each
The annotation results training initial segmentation network of structure, before obtaining the segmentation network, which comprises
The annotation results of medical image and each structure to each structure carry out left and right horizontal overturning, horizontal vertical direction
At least one of translation transformation, obtains transformed each image；
The annotation results of medical image and each structure to each structure or transformed each image are returned
At least one of one change processing, standardization processing.
7. according to the method described in claim 6, it is characterized in that, the segmentation network is full convolutional network.
8. the method according to claim 1, wherein the Morphological Identification network is multilayer convolutional network.
9. a kind of computer equipment, including memory and processor, the memory are stored with computer program, feature exists
In the step of processor realizes any one of claims 1 to 8 the method when executing the computer program.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer program
The step of method described in any item of the claim 1 to 8 is realized when being executed by processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910701918.1A CN110473208A (en) | 2019-07-31 | 2019-07-31 | Structural form method of discrimination, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910701918.1A CN110473208A (en) | 2019-07-31 | 2019-07-31 | Structural form method of discrimination, computer equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110473208A true CN110473208A (en) | 2019-11-19 |
Family
ID=68509324
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910701918.1A Pending CN110473208A (en) | 2019-07-31 | 2019-07-31 | Structural form method of discrimination, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110473208A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113782202A (en) * | 2021-08-31 | 2021-12-10 | 拓微摹心数据科技(南京)有限公司 | Transverse heart risk assessment method based on virtual annulus plane included angle of aortic valve root |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1746901A (en) * | 2005-09-19 | 2006-03-15 | 南京大学 | Computer intelligent process and processor for early lung cancer cell smear color image information |
| CN103246888A (en) * | 2012-02-03 | 2013-08-14 | 通用电气公司 | System and method for diagnosing lung disease by computer |
| CN109003672A (en) * | 2018-07-16 | 2018-12-14 | 北京睿客邦科技有限公司 | A kind of early stage of lung cancer detection classification integration apparatus and system based on deep learning |
-
2019
- 2019-07-31 CN CN201910701918.1A patent/CN110473208A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1746901A (en) * | 2005-09-19 | 2006-03-15 | 南京大学 | Computer intelligent process and processor for early lung cancer cell smear color image information |
| CN103246888A (en) * | 2012-02-03 | 2013-08-14 | 通用电气公司 | System and method for diagnosing lung disease by computer |
| CN109003672A (en) * | 2018-07-16 | 2018-12-14 | 北京睿客邦科技有限公司 | A kind of early stage of lung cancer detection classification integration apparatus and system based on deep learning |
Non-Patent Citations (4)
| Title |
|---|
| M. HOSSEIN EYBPOSH等: ""Segmentation and Classification of Cine-MR Images Using Fully Convolutional Networks and Handcrafted Features"", 《ARXIV:1709.02565V2》 * |
| SIHONG CHEN等: ""Automatic Scoring of Multiple Semantic Attributes With Multi-Task Feature Leverage:A Study on Pulmonary Nodules in CT Images"", 《IEEE TRANSACTIONS ON MEDICAL IMAGING》 * |
| VIVEK KUMAR SINGH等: ""Breast Tumor Segmentation and Shape Classification in Mammograms using Generative Adversarial and Convolutional Neural Network"", 《ARXIV:1809.01687V3》 * |
| 宋宝玉: ""基于深度学习的心脏MRI双心室自动分割与心脏病分类"", 《中国优秀硕士学位论文全文数据库·信息科技辑》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113782202A (en) * | 2021-08-31 | 2021-12-10 | 拓微摹心数据科技(南京)有限公司 | Transverse heart risk assessment method based on virtual annulus plane included angle of aortic valve root |
| CN113782202B (en) * | 2021-08-31 | 2023-08-01 | 拓微摹心数据科技(南京)有限公司 | Transverse heart risk assessment method for virtual annulus plane included angle based on aortic valve root |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111539930B (en) | Dynamic ultrasonic breast nodule real-time segmentation and identification method based on deep learning | |
| CN110197493B (en) | Fundus image blood vessel segmentation method | |
| Saikumar et al. | A novel implementation heart diagnosis system based on random forest machine learning technique. | |
| JP6058093B2 (en) | Computer-aided analysis device for medical images and computer program for medical image analysis | |
| CN110197713B (en) | Medical image processing method, device, equipment and medium | |
| CN110766730B (en) | Image registration and follow-up evaluation method, storage medium and computer equipment | |
| Nurmaini et al. | Accurate detection of septal defects with fetal ultrasonography images using deep learning-based multiclass instance segmentation | |
| CN105279759B (en) | The abdominal cavity aortic aneurysm outline dividing method constrained with reference to context information arrowband | |
| CN110400298B (en) | Method, device, equipment and medium for detecting heart clinical index | |
| Habijan et al. | Overview of the whole heart and heart chamber segmentation methods | |
| CN110838114B (en) | Pulmonary nodule detection method, device and computer storage medium | |
| de Siqueira et al. | Artificial intelligence applied to support medical decisions for the automatic analysis of echocardiogram images: A systematic review | |
| CN111260701B (en) | Multi-mode retina fundus image registration method and device | |
| CN111340825A (en) | Method and system for generating mediastinal lymph node segmentation model | |
| JP2009517163A (en) | Method, system and computer program for segmenting structure associated with reference structure in image | |
| CN111815597A (en) | Left ventricle long and short axis tangent plane extraction method and device based on CT image, computer equipment and storage medium | |
| CN111754534A (en) | CT left ventricle short axis image segmentation method and device based on deep neural network, computer equipment and storage medium | |
| CN113298773A (en) | Heart view identification and left ventricle detection device and system based on deep learning | |
| Suinesiaputra et al. | Deep learning analysis of cardiac MRI in legacy datasets: multi-ethnic study of atherosclerosis | |
| CN114693671A (en) | Lung nodule semi-automatic segmentation method, device, equipment and medium based on deep learning | |
| US11847786B2 (en) | Motion learning without labels | |
| CN110473208A (en) | Structural form method of discrimination, computer equipment and storage medium | |
| CN110992312B (en) | Medical image processing method, medical image processing device, storage medium and computer equipment | |
| CN114010227B (en) | Right ventricle characteristic information identification method and device | |
| de Melo et al. | Gradient boosting decision trees for echocardiogram images |
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 |