CN109035284A - Cardiac CT image dividing method, device, equipment and medium based on deep learning - Google Patents
Cardiac CT image dividing method, device, equipment and medium based on deep learning Download PDFInfo
- Publication number
- CN109035284A CN109035284A CN201810685558.6A CN201810685558A CN109035284A CN 109035284 A CN109035284 A CN 109035284A CN 201810685558 A CN201810685558 A CN 201810685558A CN 109035284 A CN109035284 A CN 109035284A
- Authority
- CN
- China
- Prior art keywords
- image
- cardiac
- segmentation
- heart
- net model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/174—Segmentation; Edge detection involving the use of two or more images
-
- 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
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- 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
Abstract
The present invention is applicable in technical field of medical image processing, provide a kind of cardiac CT image dividing method of deep learning, device, equipment and medium, this method comprises: when the segmentation for receiving cardiac CT image is requested, obtain the cardiac CT image of user's input, the cardiac CT image got is pre-processed, obtain corresponding pretreatment image, image segmentation is carried out to heart tissue region preset in the pretreatment image by preparatory trained V-Net model, obtain the corresponding heart tissue region segmentation image of cardiac CT image, to improve the accuracy for carrying out image segmentation to cardiac CT image, and then obtain high-precision segmented image, improve the safe coefficient of operation.
Description
Technical field
The invention belongs to technical field of medical image processing more particularly to a kind of cardiac CT images based on deep learning point
Segmentation method, device, equipment and medium.
Background technique
Heart is a part critically important in human body, heart disease have become to human life threaten biggish disease it
One.To the extraction of cardiac image interesting part be segmented in clinic study heart disease tissue play it is vital
Effect, it can assist diagnosis, reduce human error, improve medical treatment efficiency, save doctor and patient's quality time.
Computer tomography Computed Tomography, abbreviation CT) high speed development of technology constantly influences human body disease
The diagnostic mode of disease, for example, multi-layer helical and double source CT scanner can provide the fine cardiac CT image of patient, to scheme in CT
Research cardiac structure provides technical foundation as in, is widely used it in cardiac imaging.
Since the left ventricle of heart is responsible for whole body blood supply, play an important role in cardiac function and the entire heart
The region of lesion is easy in dirty, therefore left ventricle form and the exception of movement are considered as the important evidence of cardio-vascular clinical diagnosis.
To help patient to carry out the diagnosis of cranial vascular disease (Cerebrovascular Disease, abbreviation CVD), doctor is dedicated to root
Determine that left ventricular volume, myocardial wall thickness and the ventricle blood volume measured in cardiac cycle of patient (penetrate blood system according to cardiac CT image
Number) and tube wall thickening property variation, and determine myocardial wall thickness and the measurement to myocardium wall thickening rate, left ventricular volume and
Ejection fraction size all relies on the correct segmentation of myocardium of left ventricle, and therefore, myocardium of left ventricle is segmented in cardiac CT image
It has received widespread attention.
Currently, the dividing method of myocardium of left ventricle mainly include expert divide by hand, computer Interactive Segmentation and it is complete from
Dynamic segmentation.Segmentation is very high to expertise and skill requirement by hand, and unavoidably there is human error, while to magnanimity CT
It is a time-consuming and uninteresting thing that data, which carry out processing by hand, therefore, by the interactive semi-automatic segmentation of computer and complete
Automatically being segmented in the segmentation of heart CT cardiac muscle has great research significance and value.
Summary of the invention
The cardiac CT image dividing method that the purpose of the present invention is to provide a kind of based on deep learning, device, equipment and
Storage medium, it is intended to solve that a kind of effectively cardiac CT image segmentation side based on deep learning can not be provided due to the prior art
Method causes cardiac CT image to divide inaccurate problem.
On the one hand, the present invention provides a kind of cardiac CT image dividing method based on deep learning, the method includes
Following step:
When the segmentation for receiving cardiac CT image is requested, the cardiac CT image of user's input is obtained；
The cardiac CT image got is pre-processed, corresponding pretreatment image is obtained；
Image is carried out to heart tissue region preset in the pretreatment image by preparatory trained V-Net model
Segmentation, obtains the corresponding heart tissue region segmentation image of the cardiac CT image.
On the other hand, the present invention provides a kind of cardiac CT image segmenting device based on deep learning, described device packet
It includes:
CT image acquisition units, for obtaining the heart of user's input when the segmentation for receiving cardiac CT image is requested
CT image；
Image pre-processing unit obtains corresponding pre- place for pre-processing to the cardiac CT image got
Manage image；And
Segmented image acquiring unit, for being preset by preparatory trained V-Net model in the pretreatment image
Heart tissue region carry out image segmentation, obtain the corresponding heart tissue region segmentation image of the cardiac CT image.
On the other hand, the present invention also provides a kind of calculating equipment, including memory, processor and it is stored in described deposit
In reservoir and the computer program that can run on the processor, the processor are realized such as when executing the computer program
Step described in the above-mentioned cardiac CT image dividing method based on deep learning.
On the other hand, the present invention also provides a kind of computer readable storage medium, the computer readable storage mediums
It is stored with computer program, the heart CT figure as above-mentioned based on deep learning is realized when the computer program is executed by processor
The step as described in dividing method.
The present invention obtains the cardiac CT image of user's input, to obtaining when the segmentation for receiving cardiac CT image is requested
The cardiac CT image got is pre-processed, and corresponding pretreatment image is obtained, by preparatory trained V-Net model to this
Preset heart tissue region carries out image segmentation in pretreatment image, obtains the corresponding heart tissue region point of cardiac CT image
Image is cut, to improve the accuracy for carrying out image segmentation to cardiac CT image, and then high-precision segmented image is obtained, mentions
The safe coefficient of height operation.
Detailed description of the invention
Fig. 1 is the implementation process for the cardiac CT image dividing method based on deep learning that the embodiment of the present invention one provides
Figure；
Fig. 2 is the structural representation of the cardiac CT image segmenting device provided by Embodiment 2 of the present invention based on deep learning
Figure；And
Fig. 3 is the structural schematic diagram for the calculating equipment that the embodiment of the present invention three provides.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
Specific implementation of the invention is described in detail below in conjunction with specific embodiment:
Embodiment one:
Fig. 1 shows the realization stream of the cardiac CT image dividing method based on deep learning of the offer of the embodiment of the present invention one
Journey, for ease of description, only parts related to embodiments of the present invention are shown, and details are as follows:
In step s101, when the segmentation for receiving cardiac CT image is requested, the cardiac CT image of user's input is obtained.
The embodiment of the present invention is suitable for Medical Image Processing platform, system or equipment, such as personal computer, server
Deng.When the segmentation for receiving cardiac CT image is requested, the cardiac CT image of user's input is obtained, user can be from published doctor
Learn the CT cardiac image that patient is provided in the operative image of image data base or hospital's offer.The cardiac CT image is set by CT
The standby heart to patient carries out tomoscan generation.Wherein, CT scan is the X-ray beam using Accurate collimation and sensitivity pole
The cardia that high detector surrounds patient together makees profile scanning one by one, and scanning gained information is computed acquisition
The x-ray attenuation coefficient or absorption coefficient of each voxel, then these coefficients are arranged in character matrix (digital matrix),
Each number in character matrix is switched to by black to white through digital/analog converter (digital/analog converter)
The small cube of gray scales, i.e. pixel (pixel) such as not, and these pixels are arranged in a matrix, constitute cardiac CT image.
In step s 102, the cardiac CT image got is pre-processed, obtains corresponding pretreatment image.
In embodiments of the present invention, when being pre-processed to cardiac CT image, it is preferable that carried out to the cardiac CT image
Gauss and Laplce are filtered, to improve the significance of cardiac CT image minutia.
When being pre-processed to cardiac CT image, it is another preferably, the size of the cardiac CT image is normalized
Operation, to improve the subsequent accuracy being split to cardiac CT image.
In step s 103, by preparatory trained V-Net model to preset heart tissue in the pretreatment image
Region carries out image segmentation, obtains the corresponding heart tissue region segmentation image of cardiac CT image.
In embodiments of the present invention, by preparatory trained V-Net model to preset heart in the pretreatment image
Tissue regions carry out image segmentation, wherein preset heart tissue region is the tissue areas such as left ventricle and/or the right ventricle of heart
Domain.
Image is being carried out to heart tissue region preset in the pretreatment image by preparatory trained V-Net model
Before segmentation, it is preferable that one end-to-end V-Net model of building (i.e. the output image of V-Net model and V-Net model
Input picture dimension is identical), which is a kind of convolutional neural networks of V-type, including an input layer, 4 compressions
Layer, 4 decompression layers and an output layer, each compression layer are used to extract the characteristics of image of pretreatment image, and work as
The characteristics of image extracted is transmitted to next compression layer and corresponding decompression layer by preceding compression layer, so that next compression layer mentions
The characteristic information and decompression layer for taking pretreatment image deeper can be more smart according to the characteristics of image that current compression layer transmits
Restore image quasi-ly, to improve the precision of V-Net model segmented image.
When constructing an end-to-end V-Net model, it is preferable that the V-Net model further includes index linear unit, V-
All input layers, compression layer, decompression layer and every layer of output layer are correspondingly connected with an index linear unit in Net model
(Exponential Linear Unit, ELU), index linear unit is realized by ELU activation primitive, ELU activation primitive tool
There is negative value, can satisfy the requirement of zero averaging, while its assignment variation can make model training more than more gentle
It is smooth, to improve the learning efficiency of V-Net model.
When constructing an end-to-end V-Net model, it is another preferably, by each layer in V-Net model of convolution kernel
5 are dimensioned to, the convolution step-length of input layer is set as the convolution step-length of Isosorbide-5-Nitrae compression layer, 4 decompression layers and output layer
2 are disposed as, and input layer, 4 compression layers, 4 decompression layers and the corresponding feature number of channels of output layer are respectively set
It is 1,16,32,64,128,256,256,128,64 and 32, to improve the convergence rate of V-Net model.
In embodiments of the present invention, passing through preparatory trained V-Net model to the preset heart in the pretreatment image
Before dirty tissue regions carry out image segmentation, it is another preferably, by preset image processing program to heart gathered in advance
CT sample image cardiac tissue regions carry out image segmentation, the corresponding GT image in heart tissue region are obtained, according to heart CT
Sample image and GT image, are trained V-Net model, the heart tissue region segmentation that V-Net model exports after training
When the similarity degree of image and GT image meets preset threshold value, terminate the training of V-Net model, to improve V-Net model
Learning efficiency and cardiac CT image segmentation accuracy.
In embodiments of the present invention, it is preferable that preset image processing program is ITK-SNAP, and expert can pass through ITK-
SNAP sketches the contours heart tissue region (for example, left ventricle) default in cardiac CT image manually, to complete to the heart group
The calibration of tissue region generates standard picture (Ground Truth image, abbreviation GT image), to be mentioned by the GT image demarcated
The accuracy of learning efficiency and the cardiac CT image segmentation of high V-Net model.
Heart CT sample image cardiac tissue regions gathered in advance are being carried out by preset image processing program
Before image segmentation, it is preferable that operation is normalized to the size of heart CT sample image gathered in advance, to improve V-
The convergence rate of Net model training, and then improve the training effectiveness of V-Net model.
Judge V-Net model output heart tissue region segmentation image and GT image similarity degree whether meet it is pre-
If threshold value when, it is preferable that compared by broad sense Dice similarity factor (Dice Similarity Coefficient, DSC)
Similarity degree between the heart tissue region segmentation image and GT image of the segmentation of V-Net model, to improve to V-Net model
The Efficiency and accuracy of segmentation result assessment.
In embodiments of the present invention, when the segmentation for receiving cardiac CT image is requested, the heart CT of user's input is obtained
Image pre-processes the cardiac CT image got, obtains corresponding pretreatment image, passes through preparatory trained V-
Net model carries out image segmentation to heart tissue region preset in the pretreatment image, obtains the corresponding heart of cardiac CT image
Dirty tissue regions segmented image to improve the accuracy for carrying out image segmentation to cardiac CT image, and then obtains high-precision
Segmented image, improve the safe coefficient of operation.
Embodiment two:
The structure of Fig. 2 shows the provided by Embodiment 2 of the present invention cardiac CT image segmenting device based on deep learning,
For ease of description, only parts related to embodiments of the present invention are shown, including:
CT image acquisition unit 21, for obtaining the heart of user's input when the segmentation for receiving cardiac CT image is requested
Dirty CT image.
The embodiment of the present invention is suitable for Medical Image Processing platform, system or equipment, such as personal computer, server
Deng.When the segmentation for receiving cardiac CT image is requested, the cardiac CT image of user's input is obtained, user can be from published doctor
Learn the CT cardiac image that patient is provided in the operative image of image data base or hospital's offer.The cardiac CT image is set by CT
The standby heart to patient carries out tomoscan generation.Wherein, CT scan is the X-ray beam using Accurate collimation and sensitivity pole
The cardia that high detector surrounds patient together makees profile scanning one by one, and scanning gained information is computed acquisition
The x-ray attenuation coefficient or absorption coefficient of each voxel, then these coefficients are arranged in character matrix (digital matrix),
Each number in character matrix is switched to by black to white through digital/analog converter (digital/analog converter)
The small cube of gray scales, i.e. pixel (pixel) such as not, and these pixels are arranged in a matrix, constitute cardiac CT image.
Image pre-processing unit 22 obtains corresponding pretreatment for pre-processing to the cardiac CT image got
Image.
In embodiments of the present invention, when being pre-processed to cardiac CT image, it is preferable that carried out to the cardiac CT image
Gauss and Laplce are filtered, to improve the significance of cardiac CT image minutia.
When being pre-processed to cardiac CT image, it is another preferably, the size of the cardiac CT image is normalized
Operation, to improve the subsequent accuracy being split to cardiac CT image.
Image segmentation unit 23, for passing through preparatory trained V-Net model to the preset heart in the pretreatment image
Dirty tissue regions carry out image segmentation, obtain the corresponding heart tissue region segmentation image of cardiac CT image.
In embodiments of the present invention, by preparatory trained V-Net model to preset heart in the pretreatment image
Tissue regions carry out image segmentation, wherein preset heart tissue region is the tissue areas such as left ventricle and/or the right ventricle of heart
Domain.
Image is being carried out to heart tissue region preset in the pretreatment image by preparatory trained V-Net model
Before segmentation, it is preferable that one end-to-end V-Net model of building (i.e. the output image of V-Net model and V-Net model
Input picture dimension is identical), which is a kind of convolutional neural networks of V-type, including an input layer, 4 compressions
Layer, 4 decompression layers and an output layer, each compression layer are used to extract the characteristics of image of pretreatment image, and work as
The characteristics of image extracted is transmitted to next compression layer and corresponding decompression layer by preceding compression layer, so that next compression layer mentions
The characteristic information and decompression layer for taking pretreatment image deeper can be more smart according to the characteristics of image that current compression layer transmits
Restore image quasi-ly, to improve the precision of V-Net model segmented image.
When constructing an end-to-end V-Net model, it is preferable that the V-Net model further includes index linear unit, V-
All input layers, compression layer, decompression layer and every layer of output layer are correspondingly connected with an index linear unit in Net model
(Exponential Linear Unit, ELU), index linear unit is realized by ELU activation primitive, ELU activation primitive tool
There is negative value, can satisfy the requirement of zero averaging, while its assignment variation can make model training more than more gentle
It is smooth, to improve the learning efficiency of V-Net model.
When constructing an end-to-end V-Net model, it is another preferably, by each layer in V-Net model of convolution kernel
5 are dimensioned to, the convolution step-length of input layer is set as the convolution step-length of Isosorbide-5-Nitrae compression layer, 4 decompression layers and output layer
2 are disposed as, and input layer, 4 compression layers, 4 decompression layers and the corresponding feature number of channels of output layer are respectively set
It is 1,16,32,64,128,256,256,128,64 and 32, to improve the convergence rate of V-Net model.
In embodiments of the present invention, passing through preparatory trained V-Net model to the preset heart in the pretreatment image
Before dirty tissue regions carry out image segmentation, it is another preferably, by preset image processing program to heart gathered in advance
CT sample image cardiac tissue regions carry out image segmentation, the corresponding GT image in heart tissue region are obtained, according to heart CT
Sample image and GT image, are trained V-Net model, the heart tissue region segmentation that V-Net model exports after training
When the similarity degree of image and GT image meets preset threshold value, terminate the training of V-Net model, to improve V-Net model
Learning efficiency and cardiac CT image segmentation accuracy.
In embodiments of the present invention, it is preferable that preset image processing program is ITK-SNAP, and expert can pass through ITK-
SNAP sketches the contours heart tissue region (for example, left ventricle) default in cardiac CT image manually, to complete to the heart group
The calibration of tissue region generates standard picture (Ground Truth image, abbreviation GT image), to be mentioned by the GT image demarcated
The accuracy of learning efficiency and the cardiac CT image segmentation of high V-Net model.
Heart CT sample image cardiac tissue regions gathered in advance are being carried out by preset image processing program
Before image segmentation, it is preferable that operation is normalized to the size of heart CT sample image gathered in advance, to improve V-
The convergence rate of Net model training, and then improve the training effectiveness of V-Net model.
Judge V-Net model output heart tissue region segmentation image and GT image similarity degree whether meet it is pre-
If threshold value when, it is preferable that compared by broad sense Dice similarity factor (Dice Similarity Coefficient, DSC)
Similarity degree between the heart tissue region segmentation image and GT image of the segmentation of V-Net model, to improve to V-Net model
The Efficiency and accuracy of segmentation result assessment.
It is therefore preferred that the cardiac CT image segmenting device based on deep learning of the embodiment of the present invention further include:
GT image acquisition unit, for passing through preset image processing program in heart CT sample image gathered in advance
Heart tissue region carries out image segmentation, obtains the corresponding GT image of heart CT sample image cardiac tissue regions；
Model training unit, for being trained to V-Net model according to heart CT sample image and GT image；And
Training end unit, heart tissue region segmentation image and GT image for V-Net model output after training
Similarity degree when meeting preset threshold value, terminate the training of V-Net model.
In embodiments of the present invention, each unit of the cardiac CT image segmenting device based on deep learning can be by corresponding hard
Part or software unit realize that each unit can be independent soft and hardware unit, also can integrate as a soft and hardware unit,
This is not to limit the present invention.
Embodiment three:
Fig. 3 shows the structure of the calculating equipment of the offer of the embodiment of the present invention three, for ease of description, illustrates only and this
The relevant part of inventive embodiments.
The calculating equipment 3 of the embodiment of the present invention includes processor 30, memory 31 and is stored in memory 31 and can
The computer program 32 run on processor 30.The processor 30 realizes above-mentioned deep learning when executing computer program 32
Step in cardiac CT image dividing method embodiment, such as step S101 to S103 shown in FIG. 1.Alternatively, processor 30 is held
The function of each unit in above-mentioned each Installation practice, such as the function of unit 21 to 23 shown in Fig. 2 are realized when row computer program 32
Energy.
In embodiments of the present invention, when the segmentation for receiving cardiac CT image is requested, the heart CT of user's input is obtained
Image pre-processes the cardiac CT image got, obtains corresponding pretreatment image, passes through preparatory trained V-
Net model carries out image segmentation to heart tissue region preset in the pretreatment image, obtains the corresponding heart of cardiac CT image
Dirty tissue regions segmented image to improve the accuracy for carrying out image segmentation to cardiac CT image, and then obtains high-precision
Segmented image, improve the safe coefficient of operation.
The calculating equipment of the embodiment of the present invention can be personal computer, server.Processor 30 is held in the calculating equipment 3
The step of realizing when realizing the cardiac CT image dividing method of deep learning when row computer program 32 can refer to preceding method reality
The description of example is applied, details are not described herein.
Example IV:
In embodiments of the present invention, a kind of computer readable storage medium is provided, which deposits
Computer program is contained, which realizes the cardiac CT image dividing method of above-mentioned deep learning when being executed by processor
Step in embodiment, for example, step S101 to S103 shown in FIG. 1.Alternatively, real when the computer program is executed by processor
The function of each unit in existing above-mentioned each Installation practice, such as the function of unit 21 to 23 shown in Fig. 2.
In embodiments of the present invention, when the segmentation for receiving cardiac CT image is requested, the heart CT of user's input is obtained
Image pre-processes the cardiac CT image got, obtains corresponding pretreatment image, passes through preparatory trained V-
Net model carries out image segmentation to heart tissue region preset in the pretreatment image, obtains the corresponding heart of cardiac CT image
Dirty tissue regions segmented image to improve the accuracy for carrying out image segmentation to cardiac CT image, and then obtains high-precision
Segmented image, improve the safe coefficient of operation.
The computer readable storage medium of the embodiment of the present invention may include can carry computer program code any
Entity or device, recording medium, for example, the memories such as ROM/RAM, disk, CD, flash memory.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Made any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention within mind and principle.
Claims (10)
1. a kind of cardiac CT image dividing method based on deep learning, which is characterized in that the method includes the following steps:
When the segmentation for receiving cardiac CT image is requested, the cardiac CT image of user's input is obtained；
The cardiac CT image got is pre-processed, corresponding pretreatment image is obtained；
Image point is carried out to heart tissue region preset in the pretreatment image by preparatory trained V-Net model
It cuts, obtains the corresponding heart tissue region segmentation image of the cardiac CT image.
2. the method as described in claim 1, which is characterized in that before the step of obtaining the cardiac CT image of user's input, institute
State method further include:
Figure is carried out to heart tissue region described in heart CT sample image gathered in advance by preset image processing program
As segmentation, the corresponding GT image in heart tissue region described in the heart CT sample image is obtained；
According to the heart CT sample image and the GT image, the V-Net model is trained；
The similarity degree of the heart tissue region segmentation image and the GT image of V-Net model output after training
When meeting preset threshold value, terminate the training of the V-Net model.
3. method according to claim 2, which is characterized in that the preset image processing program is ITK-SNAP.
4. the method as described in claim 1, which is characterized in that the V-Net model include an input layer, 4 compression layers,
4 decompression layers and an output layer, each layer of convolution kernel size is 5 in the V-Net model, the volume of the input layer
Product step-length is 1, and the convolution step-length of the compression layer, the decompression layer and the output layer is 2.
5. a kind of cardiac CT image segmenting device based on deep learning, which is characterized in that described device includes:
CT image acquisition unit, for obtaining the heart CT figure of user's input when the segmentation for receiving cardiac CT image is requested
Picture；
Image pre-processing unit obtains corresponding pretreatment figure for pre-processing to the cardiac CT image got
Picture；And
Image segmentation unit, for passing through preparatory trained V-Net model to heart group preset in the pretreatment image
Tissue region carries out image segmentation, obtains the corresponding heart tissue region segmentation image of the cardiac CT image.
6. device as claimed in claim 5, which is characterized in that described device further include:
GT image acquisition unit, for passing through preset image processing program to described in heart CT sample image gathered in advance
Heart tissue region carries out image segmentation, obtains the corresponding GT figure in heart tissue region described in the heart CT sample image
Picture；
Model training unit, for being instructed to the V-Net model according to the heart CT sample image and the GT image
Practice；And
Training end unit, the heart tissue region segmentation image and institute for the V-Net model output described after training
When stating the similarity degree of GT image and meeting preset threshold value, terminate the training of the V-Net model.
7. device as claimed in claim 6, which is characterized in that the preset image processing program is ITK-SNAP.
8. device as claimed in claim 5, which is characterized in that the V-Net model include an input layer, 4 compression layers,
4 decompression layers and an output layer, each layer of convolution kernel size is 5 in the V-Net model, the volume of the input layer
Product step-length is 1, and the convolution step-length of the compression layer, the decompression layer and the output layer is 2.
9. a kind of calculating equipment, including memory, processor and storage are in the memory and can be on the processor
The computer program of operation, which is characterized in that the processor realizes such as Claims 1-4 when executing the computer program
The step of any one the method.
10. a kind of computer readable storage medium, the computer-readable recording medium storage has computer program, and feature exists
In when the computer program is executed by processor the step of any one of such as Claims 1-4 of realization the method.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810685558.6A CN109035284B (en) | 2018-06-28 | 2018-06-28 | Heart CT image segmentation method, device, equipment and medium based on deep learning |
| PCT/CN2019/078534 WO2020001086A1 (en) | 2018-06-28 | 2019-03-18 | Deep learning-based cardiac ct image segmentation method and apparatus, device, and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810685558.6A CN109035284B (en) | 2018-06-28 | 2018-06-28 | Heart CT image segmentation method, device, equipment and medium based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109035284A true CN109035284A (en) | 2018-12-18 |
| CN109035284B CN109035284B (en) | 2022-05-06 |
Family
ID=65520688
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810685558.6A Active CN109035284B (en) | 2018-06-28 | 2018-06-28 | Heart CT image segmentation method, device, equipment and medium based on deep learning |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN109035284B (en) |
| WO (1) | WO2020001086A1 (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109598734A (en) * | 2018-12-29 | 2019-04-09 | 上海联影智能医疗科技有限公司 | The method and system of heart and lobe of the lung segmentation |
| CN109949309A (en) * | 2019-03-18 | 2019-06-28 | 安徽紫薇帝星数字科技有限公司 | A kind of CT image for liver dividing method based on deep learning |
| CN110610487A (en) * | 2019-08-29 | 2019-12-24 | 上海杏脉信息科技有限公司 | Method and device for measuring increase of heart shadow |
| WO2020001086A1 (en) * | 2018-06-28 | 2020-01-02 | 深圳先进技术研究院 | Deep learning-based cardiac ct image segmentation method and apparatus, device, and medium |
| CN111626972A (en) * | 2020-06-02 | 2020-09-04 | 上海鹰瞳医疗科技有限公司 | CT image reconstruction method, model training method and equipment |
| CN112070809A (en) * | 2020-07-22 | 2020-12-11 | 中国科学院苏州生物医学工程技术研究所 | Accurate diagnosis system of pancreatic cancer based on two time formation of image of PET/CT |
| CN113205509A (en) * | 2021-05-24 | 2021-08-03 | 山东省人工智能研究院 | Blood vessel plaque CT image segmentation method based on position convolution attention network |
| CN113240661A (en) * | 2021-05-31 | 2021-08-10 | 平安科技（深圳）有限公司 | Deep learning-based lumbar vertebra analysis method, device, equipment and storage medium |
| CN113450337A (en) * | 2021-07-07 | 2021-09-28 | 沈阳先进医疗设备技术孵化中心有限公司 | Evaluation method and device for hydrops in pericardial cavity, electronic device and storage medium |
| US11436720B2 (en) | 2018-12-28 | 2022-09-06 | Shanghai United Imaging Intelligence Co., Ltd. | Systems and methods for generating image metric |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111768391B (en) * | 2020-07-01 | 2024-04-05 | 杭州脉流科技有限公司 | Full-automatic heart function analysis method, device, computer equipment and storage medium based on CT image |
| CN115170591A (en) * | 2022-09-02 | 2022-10-11 | 湖南红普创新科技发展有限公司 | Method and device for acquiring lesion area image and related equipment |
| CN115329952B (en) * | 2022-10-12 | 2023-01-20 | 深圳比特微电子科技有限公司 | Model compression method and device and readable storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080181479A1 (en) * | 2002-06-07 | 2008-07-31 | Fuxing Yang | System and method for cardiac imaging |
| CN105574859A (en) * | 2015-12-14 | 2016-05-11 | 中国科学院深圳先进技术研究院 | Liver tumor segmentation method and device based on CT (Computed Tomography) image |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107784647B (en) * | 2017-09-29 | 2021-03-09 | 华侨大学 | Liver and tumor segmentation method and system based on multitask deep convolutional network |
| CN107705305A (en) * | 2017-10-20 | 2018-02-16 | 上海联影医疗科技有限公司 | A kind of method and device of Medical Image Processing |
| CN109615636B (en) * | 2017-11-03 | 2020-06-12 | 杭州依图医疗技术有限公司 | Blood vessel tree construction method and device in lung lobe segment segmentation of CT (computed tomography) image |
| CN108010021B (en) * | 2017-11-30 | 2021-12-10 | 上海联影医疗科技股份有限公司 | Medical image processing system and method |
| CN109035284B (en) * | 2018-06-28 | 2022-05-06 | 深圳先进技术研究院 | Heart CT image segmentation method, device, equipment and medium based on deep learning |
-
2018
- 2018-06-28 CN CN201810685558.6A patent/CN109035284B/en active Active
-
2019
- 2019-03-18 WO PCT/CN2019/078534 patent/WO2020001086A1/en active Application Filing
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080181479A1 (en) * | 2002-06-07 | 2008-07-31 | Fuxing Yang | System and method for cardiac imaging |
| CN105574859A (en) * | 2015-12-14 | 2016-05-11 | 中国科学院深圳先进技术研究院 | Liver tumor segmentation method and device based on CT (Computed Tomography) image |
Non-Patent Citations (4)
| Title |
|---|
| CASTILLO L S等: "Volumetric multimodality neural network for brain tumor segmentation", 《13TH INTERNATIONAL CONFERENCE ON MEDICAL INFORMATION PROCESSING AND ANALYSIS. INTERNATIONAL SOCIETY FOR OPTICS AND PHOTONICS, 2017》 * |
| MILLETARI F等: "V-net: Fully convolutional neural networks for volumetric medical image segmentation", 《2016 FOURTH INTERNATIONAL CONFERENCE ON 3D VISION (3DV). IEEE》 * |
| TANG H等: "Segmentation of anatomical structures in cardiac CTA using multi-label V-Net", 《MEDICAL IMAGING 2018: IMAGE PROCESSING. INTERNATIONAL SOCIETY FOR OPTICS AND PHOTONICS》 * |
| 张品等: "基于改进C-V模型的肾脏CT图像分割方法", 《光电子.激光》 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020001086A1 (en) * | 2018-06-28 | 2020-01-02 | 深圳先进技术研究院 | Deep learning-based cardiac ct image segmentation method and apparatus, device, and medium |
| US11436720B2 (en) | 2018-12-28 | 2022-09-06 | Shanghai United Imaging Intelligence Co., Ltd. | Systems and methods for generating image metric |
| CN109598734A (en) * | 2018-12-29 | 2019-04-09 | 上海联影智能医疗科技有限公司 | The method and system of heart and lobe of the lung segmentation |
| CN109949309A (en) * | 2019-03-18 | 2019-06-28 | 安徽紫薇帝星数字科技有限公司 | A kind of CT image for liver dividing method based on deep learning |
| CN110610487A (en) * | 2019-08-29 | 2019-12-24 | 上海杏脉信息科技有限公司 | Method and device for measuring increase of heart shadow |
| CN111626972A (en) * | 2020-06-02 | 2020-09-04 | 上海鹰瞳医疗科技有限公司 | CT image reconstruction method, model training method and equipment |
| CN111626972B (en) * | 2020-06-02 | 2023-06-02 | 上海鹰瞳医疗科技有限公司 | CT image reconstruction method, model training method and equipment |
| CN112070809A (en) * | 2020-07-22 | 2020-12-11 | 中国科学院苏州生物医学工程技术研究所 | Accurate diagnosis system of pancreatic cancer based on two time formation of image of PET/CT |
| CN112070809B (en) * | 2020-07-22 | 2024-01-26 | 中国科学院苏州生物医学工程技术研究所 | Pancreatic cancer accurate diagnosis system based on PET/CT double-time imaging |
| CN113205509A (en) * | 2021-05-24 | 2021-08-03 | 山东省人工智能研究院 | Blood vessel plaque CT image segmentation method based on position convolution attention network |
| CN113205509B (en) * | 2021-05-24 | 2021-11-09 | 山东省人工智能研究院 | Blood vessel plaque CT image segmentation method based on position convolution attention network |
| CN113240661B (en) * | 2021-05-31 | 2023-09-26 | 平安科技（深圳）有限公司 | Deep learning-based lumbar vertebra bone analysis method, device, equipment and storage medium |
| CN113240661A (en) * | 2021-05-31 | 2021-08-10 | 平安科技（深圳）有限公司 | Deep learning-based lumbar vertebra analysis method, device, equipment and storage medium |
| CN113450337A (en) * | 2021-07-07 | 2021-09-28 | 沈阳先进医疗设备技术孵化中心有限公司 | Evaluation method and device for hydrops in pericardial cavity, electronic device and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2020001086A1 (en) | 2020-01-02 |
| CN109035284B (en) | 2022-05-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109035284A (en) | Cardiac CT image dividing method, device, equipment and medium based on deep learning | |
| US11398029B2 (en) | Systems and methods for medical acquisition processing and machine learning for anatomical assessment | |
| US11501485B2 (en) | System and method for image-based object modeling using multiple image acquisitions or reconstructions | |
| US9514530B2 (en) | Systems and methods for image-based object modeling using multiple image acquisitions or reconstructions | |
| CN109598722B (en) | Image analysis method based on recurrent neural network | |
| KR101902883B1 (en) | A method for analyzing plaque in a computed tomography image and an apparatus thereof | |
| JP7059298B2 (en) | Myocardial CT perfusion image synthesis | |
| KR101978317B1 (en) | Ct image database-based cardiac image segmentation method and an apparatus thereof | |
| US11823384B2 (en) | CT image generation method for attenuation correction of pet images | |
| CN109961419B (en) | Correction information acquisition method for attenuation correction of PET activity distribution image | |
| Nguyen et al. | 3D Unet generative adversarial network for attenuation correction of SPECT images | |
| Shi et al. | A Virtual Monochromatic Imaging Method for Spectral CT Based on Wasserstein Generative Adversarial Network With a Hybrid Loss. | |
| CN116168099A (en) | Medical image reconstruction method and device and nonvolatile storage medium | |
| Spiesberger et al. | Processing of medical image sequences | |
| US20220044454A1 (en) | Deep reinforcement learning for computer assisted reading and analysis | |
| Qin et al. | Artificial intelligence: The future for multimodality imaging of right ventricle | |
| WO2023131061A1 (en) | Systems and methods for positron emission computed tomography image reconstruction | |
| da Silva Corado | Echocardiography Automatic Image Quality Enhancement Using Generative Adversarial Networks | |
| CN117788624A (en) | Scattering correction method and system for PET image | |
| CN113643285A (en) | Myocardial bulls-eye diagram analysis method, system and medium based on artificial neural network | |
| Abboud | Segmenting the Right Ventricle Cavity from 4D Echocardiography Images for Stroke Volume Measurement | |
| Stetten | Automated identification and measurement of cardiac anatomy via statistical analysis of medial primitives | |
| Tobon-Gomez et al. | Systolic and diastolic assessment by 3D-ASM segmentation of gated-SPECT Studies: a comparison with MRI |
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 |