CN109035284B - Heart CT image segmentation method, device, equipment and medium based on deep learning - Google Patents
Heart CT image segmentation method, device, equipment and medium based on deep learning Download PDFInfo
- Publication number
- CN109035284B CN109035284B CN201810685558.6A CN201810685558A CN109035284B CN 109035284 B CN109035284 B CN 109035284B CN 201810685558 A CN201810685558 A CN 201810685558A CN 109035284 B CN109035284 B CN 109035284B
- Authority
- CN
- China
- Prior art keywords
- image
- heart
- segmentation
- cardiac
- 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.)
- Active
Links
Images
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 invention is suitable for the technical field of medical image processing, and provides a heart CT image segmentation method, a device, equipment and a medium for deep learning, wherein the method comprises the following steps: when a heart CT image segmentation request is received, a heart CT image input by a user is obtained, the obtained heart CT image is preprocessed to obtain a corresponding preprocessed image, a preset heart tissue region in the preprocessed image is segmented through a pre-trained V-Net model to obtain a heart tissue region segmentation image corresponding to the heart CT image, so that the accuracy of image segmentation on the heart CT image is improved, a high-precision segmentation image is obtained, and the safety degree of an operation is improved.
Description
Technical Field
The invention belongs to the technical field of medical image processing, and particularly relates to a heart CT image segmentation method, device, equipment and medium based on deep learning.
Background
The heart is an important part of the human body, and heart diseases have become one of the diseases which are more life-threatening to human beings. The extraction and segmentation of the interested part of the heart image play an important role in the clinical medicine research of the heart lesion tissue, and can assist the diagnosis of doctors, reduce human errors, improve the medical efficiency and save the precious time of doctors and patients.
Computed Tomography (CT) technology, for example, a multi-slice helical and dual-source CT scanner can provide a refined cardiac CT image of a patient, and provide a technical basis for studying cardiac structures in the CT image, so that the CT scanner is widely applied to cardiac imaging.
Since the left ventricle of the heart is responsible for supplying blood to the whole body, plays an important role in cardiac function, and is also an easily diseased area in the whole heart, abnormalities in left ventricle morphology and motion are regarded as an important basis for cardiovascular clinical diagnosis. To assist the patient in diagnosing Cerebrovascular Disease (CVD), doctors are dedicated to determining the left ventricular volume and myocardial wall thickness of the patient from the cardiac CT image and measuring the change of ventricular blood volume (ejection fraction) and wall thickening property in the cardiac cycle, and the determination of the myocardial wall thickness and the measurement of the myocardial wall thickening rate, the left ventricular volume and the ejection fraction are all dependent on the correct segmentation of the left ventricular myocardium, so the segmentation of the left ventricular myocardium is of great interest in the cardiac CT image.
At present, the segmentation method of the left ventricular myocardium mainly comprises expert manual segmentation, computer interactive segmentation and full-automatic segmentation. The manual segmentation has high requirements on expert knowledge and experience, human errors inevitably exist, and meanwhile, the manual processing of massive CT data is time-consuming and tedious, so that interactive semi-automatic segmentation and full-automatic segmentation by means of a computer have great research significance and value in the segmentation of cardiac CT myocardium.
Disclosure of Invention
The invention aims to provide a heart CT image segmentation method, a heart CT image segmentation device, heart CT image segmentation equipment and a heart CT image storage medium based on deep learning, and aims to solve the problem that in the prior art, an effective heart CT image segmentation method based on deep learning cannot be provided, so that the heart CT image segmentation is inaccurate.
In one aspect, the present invention provides a cardiac CT image segmentation method based on deep learning, including the following steps:
when a segmentation request of a cardiac CT image is received, acquiring the cardiac CT image input by a user;
preprocessing the acquired cardiac CT image to obtain a corresponding preprocessed image;
and carrying out image segmentation on a preset heart tissue region in the preprocessed image through a pre-trained V-Net model to obtain a heart tissue region segmentation image corresponding to the heart CT image.
In another aspect, the present invention provides a cardiac CT image segmentation apparatus based on deep learning, including:
the CT image acquisition unit is used for acquiring a cardiac CT image input by a user when a segmentation request of the cardiac CT image is received;
the image preprocessing unit is used for preprocessing the acquired cardiac CT image to acquire a corresponding preprocessed image; and
and the segmented image acquisition unit is used for carrying out image segmentation on a preset heart tissue region in the preprocessed image through a pre-trained V-Net model to obtain a heart tissue region segmented image corresponding to the heart CT image.
In another aspect, the present invention further provides a computing device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the deep learning-based cardiac CT image segmentation method when executing the computer program.
In another aspect, the present invention further provides a computer-readable storage medium, which stores a computer program, which when executed by a processor implements the steps of the above-mentioned cardiac CT image segmentation method based on deep learning.
According to the invention, when a heart CT image segmentation request is received, a heart CT image input by a user is obtained, the obtained heart CT image is preprocessed to obtain a corresponding preprocessed image, and a preset heart tissue region in the preprocessed image is segmented by a pre-trained V-Net model to obtain a heart tissue region segmentation image corresponding to the heart CT image, so that the accuracy of image segmentation on the heart CT image is improved, a high-precision segmentation image is obtained, and the safety degree of an operation is improved.
Drawings
Fig. 1 is a flowchart illustrating an implementation of a cardiac CT image segmentation method based on deep learning according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of a deep learning-based cardiac CT image segmentation apparatus according to a second embodiment of the present invention; and
fig. 3 is a schematic structural diagram of a computing device according to a third embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The following detailed description of specific implementations of the present invention is provided in conjunction with specific embodiments:
the first embodiment is as follows:
fig. 1 shows a flow of implementing the method for cardiac CT image segmentation based on deep learning according to an embodiment of the present invention, and for convenience of description, only the parts related to the embodiment of the present invention are shown, which is detailed as follows:
in step S101, when a cardiac CT image segmentation request is received, a cardiac CT image input by a user is acquired.
The embodiment of the invention is suitable for medical image processing platforms, systems or equipment, such as personal computers, servers and the like. When a cardiac CT image segmentation request is received, a cardiac CT image input by a user is acquired, and the user can acquire a CT cardiac image of a patient from a published medical image database or a surgery image provided by a hospital. The cardiac CT image is generated by tomographic scanning of the heart of the patient by a CT apparatus. The CT scanning is to make one-by-one cross-section scanning around the heart part of the patient by utilizing an X-ray beam with precise collimation and a detector with extremely high sensitivity, the obtained information of the scanning is calculated to obtain the X-ray attenuation coefficient or absorption coefficient of each voxel, then the coefficients are arranged into a digital matrix (digital matrix), each digital in the digital matrix is converted into small blocks with different gray scales from black to white, namely pixels (pixels), through a digital/analog converter, and the pixels are arranged according to the matrix to form the heart CT image.
In step S102, the acquired cardiac CT image is preprocessed to obtain a corresponding preprocessed image.
In the embodiment of the present invention, when the cardiac CT image is preprocessed, it is preferable to perform gaussian and laplacian filtering on the cardiac CT image, so as to improve the significance of detail features of the cardiac CT image.
In the preprocessing of the cardiac CT image, it is further preferable that the size of the cardiac CT image is normalized, so as to improve the accuracy of the subsequent segmentation of the cardiac CT image.
In step S103, a heart tissue region segmentation image corresponding to the cardiac CT image is obtained by performing image segmentation on a preset heart tissue region in the preprocessed image through a pre-trained V-Net model.
In the embodiment of the invention, the preset heart tissue region in the preprocessed image is subjected to image segmentation through a pre-trained V-Net model, wherein the preset heart tissue region is a tissue region of the heart, such as the left ventricle and/or the right ventricle.
Before image segmentation is carried out on the preset heart tissue region in the preprocessed image through a pre-trained V-Net model, an end-to-end V-Net model is preferably constructed (namely an output image of the V-Net model and an input image of the V-Net model have the same dimension), the V-Net model is a V-shaped convolutional neural network, which comprises an input layer, 4 compression layers, 4 decompression layers and an output layer, wherein each compression layer is used for extracting the image characteristics of a preprocessed image, and the current compression layer transfers the extracted image features to the next compression layer and the corresponding decompression layer, so that the next compression layer extracts feature information of deeper layers of the preprocessed image, and the decompression layer can restore the image more accurately according to the image characteristics transmitted by the current compression layer, so that the image segmentation accuracy of the V-Net model is improved.
When an end-to-end V-Net model is constructed, preferably, the V-Net model further comprises an Exponential Linear Unit (ELU), wherein each layer of all input layers, compression layers, decompression layers and output layers in the V-Net model is correspondingly connected with one Exponential Linear Unit (ELU), the Exponential Linear Unit is realized through an ELU activation function, the ELU activation function has a negative value and can meet the requirement of zero averaging, and meanwhile, the assignment change of the ELU activation function is relatively gentle, so that the model training is smoother, and the learning efficiency of the V-Net model is improved.
When constructing an end-to-end V-Net model, it is further preferable that the convolution kernel size of each layer in the V-Net model is set to 5, the convolution step size of the input layer is set to 1, the convolution step size of each of the compression layer, the decompression layer and the output layer is set to 2, and the number of feature channels corresponding to the input layer, the compression layer, the decompression layer and the output layer is set to 1, 16, 32, 64, 128, 256, 128, 64 and 32, respectively, so as to improve the convergence speed of the V-Net model.
In the embodiment of the present invention, before image segmentation is performed on a preset heart tissue region in the preprocessed image through a pre-trained V-Net model, it is further preferable that image segmentation is performed on the heart tissue region in a pre-acquired heart CT sample image through a preset image processing program to obtain a GT image corresponding to the heart tissue region, the V-Net model is trained according to the heart CT sample image and the GT image, and when a similarity degree between a heart tissue region segmentation image output by the trained V-Net model and the GT image satisfies a preset threshold, training of the V-Net model is ended, so as to improve learning efficiency of the V-Net model and accuracy of heart CT image segmentation.
In the embodiment of the present invention, preferably, the preset image processing program is ITK-SNAP, and an expert may manually outline a preset cardiac tissue region (for example, the left ventricle) in the cardiac CT image through ITK-SNAP to complete calibration of the cardiac tissue region, so as to generate a standard image (a group route image, which is referred to as a GT image for short), thereby improving the learning efficiency of the V-Net model and the accuracy of cardiac CT image segmentation through the calibrated GT image.
Before image segmentation is carried out on a cardiac tissue region in a cardiac CT sample image acquired in advance through a preset image processing program, preferably, normalization operation is carried out on the size of the cardiac CT sample image acquired in advance, so that the convergence speed of V-Net model training is improved, and further the training efficiency of the V-Net model is improved.
When judging whether the Similarity degree between the heart tissue area segmentation image output by the V-Net model and the GT image meets a preset threshold value, preferably, the Similarity degree between the heart tissue area segmentation image segmented by the V-Net model and the GT image is compared through a generalized Dice Similarity Coefficient (DSC), so that the efficiency and the accuracy of the evaluation of the V-Net model segmentation result are improved.
In the embodiment of the invention, when a heart CT image segmentation request is received, a heart CT image input by a user is acquired, the acquired heart CT image is preprocessed to acquire a corresponding preprocessed image, and a pre-trained V-Net model is used for carrying out image segmentation on a preset heart tissue region in the preprocessed image to acquire a heart tissue region segmentation image corresponding to the heart CT image, so that the accuracy of image segmentation on the heart CT image is improved, a high-precision segmentation image is acquired, and the safety degree of an operation is improved.
Example two:
fig. 2 illustrates a structure of a cardiac CT image segmentation apparatus based on deep learning according to a second embodiment of the present invention, and for convenience of description, only the portions related to the second embodiment of the present invention are illustrated, which include:
a CT image obtaining unit 21, configured to obtain a cardiac CT image input by a user when a segmentation request of the cardiac CT image is received.
The embodiment of the invention is suitable for medical image processing platforms, systems or equipment, such as personal computers, servers and the like. When a cardiac CT image segmentation request is received, a cardiac CT image input by a user is acquired, and the user can acquire a CT cardiac image of a patient from a published medical image database or a surgery image provided by a hospital. The cardiac CT image is generated by tomographic scanning of the heart of the patient by a CT apparatus. The CT scanning is to make one-by-one cross-section scanning around the heart part of the patient by utilizing an X-ray beam with precise collimation and a detector with extremely high sensitivity, the obtained information of the scanning is calculated to obtain the X-ray attenuation coefficient or absorption coefficient of each voxel, then the coefficients are arranged into a digital matrix (digital matrix), each digital in the digital matrix is converted into small blocks with different gray scales from black to white, namely pixels (pixels), through a digital/analog converter, and the pixels are arranged according to the matrix to form the heart CT image.
And the image preprocessing unit 22 is configured to preprocess the acquired cardiac CT image to obtain a corresponding preprocessed image.
In the embodiment of the present invention, when the cardiac CT image is preprocessed, it is preferable to perform gaussian and laplacian filtering on the cardiac CT image, so as to improve the significance of detail features of the cardiac CT image.
In the preprocessing of the cardiac CT image, it is further preferable that the size of the cardiac CT image is normalized, so as to improve the accuracy of the subsequent segmentation of the cardiac CT image.
And the image segmentation unit 23 is configured to perform image segmentation on the preset cardiac tissue region in the preprocessed image through a pre-trained V-Net model, so as to obtain a cardiac tissue region segmentation image corresponding to the cardiac CT image.
In the embodiment of the invention, the preset heart tissue region in the preprocessed image is subjected to image segmentation through a pre-trained V-Net model, wherein the preset heart tissue region is a tissue region of the heart, such as the left ventricle and/or the right ventricle.
Before image segmentation is carried out on the preset heart tissue region in the preprocessed image through a pre-trained V-Net model, an end-to-end V-Net model is preferably constructed (namely an output image of the V-Net model and an input image of the V-Net model have the same dimension), the V-Net model is a V-shaped convolutional neural network, which comprises an input layer, 4 compression layers, 4 decompression layers and an output layer, wherein each compression layer is used for extracting the image characteristics of a preprocessed image, and the current compression layer transfers the extracted image features to the next compression layer and the corresponding decompression layer, so that the next compression layer extracts feature information of deeper layers of the preprocessed image, and the decompression layer can restore the image more accurately according to the image characteristics transmitted by the current compression layer, so that the image segmentation accuracy of the V-Net model is improved.
When an end-to-end V-Net model is constructed, preferably, the V-Net model further comprises an Exponential Linear Unit (ELU), wherein each layer of all input layers, compression layers, decompression layers and output layers in the V-Net model is correspondingly connected with one Exponential Linear Unit (ELU), the Exponential Linear Unit is realized through an ELU activation function, the ELU activation function has a negative value and can meet the requirement of zero averaging, and meanwhile, the assignment change of the ELU activation function is relatively gentle, so that the model training is smoother, and the learning efficiency of the V-Net model is improved.
When constructing an end-to-end V-Net model, it is further preferable that the convolution kernel size of each layer in the V-Net model is set to 5, the convolution step size of the input layer is set to 1, the convolution step size of each of the 4 compression layers, the convolution step size of each of the 4 decompression layers, and the convolution step size of each of the output layers is set to 2, and the numbers of feature channels corresponding to the input layer, the 4 compression layers, the 4 decompression layers, and the output layers are set to 1, 16, 32, 64, 128, 256, 128, 64, and 32, respectively, so as to improve the convergence speed of the V-Net model.
In the embodiment of the present invention, before image segmentation is performed on a preset cardiac tissue region in the preprocessed image through a pre-trained V-Net model, it is further preferable that image segmentation is performed on the cardiac tissue region in a pre-acquired cardiac CT sample image through a preset image processing program to obtain a GT image corresponding to the cardiac tissue region, the V-Net model is trained according to the cardiac CT sample image and the GT image, and when a similarity between a cardiac tissue region segmentation image output by the trained V-Net model and the GT image satisfies a preset threshold, training of the V-Net model is ended, so that learning efficiency of the V-Net model and accuracy of cardiac CT image segmentation are improved.
In the embodiment of the present invention, preferably, the preset image processing program is ITK-SNAP, and an expert may manually outline a preset cardiac tissue region (for example, the left ventricle) in the cardiac CT image through ITK-SNAP to complete calibration of the cardiac tissue region, so as to generate a standard image (a group route image, which is referred to as a GT image for short), thereby improving the learning efficiency of the V-Net model and the accuracy of cardiac CT image segmentation through the calibrated GT image.
Before image segmentation is carried out on a cardiac tissue region in a cardiac CT sample image acquired in advance through a preset image processing program, preferably, normalization operation is carried out on the size of the cardiac CT sample image acquired in advance, so that the convergence speed of V-Net model training is improved, and further the training efficiency of the V-Net model is improved.
When judging whether the Similarity degree between the heart tissue area segmentation image output by the V-Net model and the GT image meets a preset threshold value, preferably, the Similarity degree between the heart tissue area segmentation image segmented by the V-Net model and the GT image is compared through a generalized Dice Similarity Coefficient (DSC), so that the efficiency and the accuracy of the evaluation of the V-Net model segmentation result are improved.
Therefore, preferably, the cardiac CT image segmentation apparatus based on deep learning according to the embodiment of the present invention further includes:
the GT image acquisition unit is used for carrying out image segmentation on a heart tissue area in a heart CT sample image acquired in advance through a preset image processing program to obtain a GT image corresponding to the heart tissue area in the heart CT sample image;
the model training unit is used for training the V-Net model according to the heart CT sample image and the GT image; and
and the training ending unit is used for ending the training of the V-Net model when the similarity degree of the heart tissue region segmentation image output by the trained V-Net model and the GT image meets a preset threshold value.
In the embodiment of the present invention, each unit of the cardiac CT image segmentation apparatus based on deep learning may be implemented by corresponding hardware or software unit, and each unit may be an independent software or hardware unit, or may be integrated into a software or hardware unit, which is not limited herein.
Example three:
fig. 3 shows a structure of a computing device provided in a third embodiment of the present invention, and for convenience of explanation, only a part related to the third embodiment of the present invention is shown.
The computing device 3 of an embodiment of the invention comprises a processor 30, a memory 31 and a computer program 32 stored in the memory 31 and executable on the processor 30. The processor 30 executes the computer program 32 to implement the steps in the above-mentioned deep learning cardiac CT image segmentation method embodiment, such as the steps S101 to S103 shown in fig. 1. Alternatively, the processor 30, when executing the computer program 32, implements the functions of the units in the above-described device embodiments, such as the functions of the units 21 to 23 shown in fig. 2.
In the embodiment of the invention, when a heart CT image segmentation request is received, a heart CT image input by a user is acquired, the acquired heart CT image is preprocessed to acquire a corresponding preprocessed image, and a pre-trained V-Net model is used for carrying out image segmentation on a preset heart tissue region in the preprocessed image to acquire a heart tissue region segmentation image corresponding to the heart CT image, so that the accuracy of image segmentation on the heart CT image is improved, a high-precision segmentation image is acquired, and the safety degree of an operation is improved.
The computing equipment of the embodiment of the invention can be a personal computer and a server. The steps of the method for segmenting cardiac CT images for deep learning implemented by the processor 30 in the computing device 3 when executing the computer program 32 can refer to the description of the foregoing method embodiments, and are not repeated herein.
Example four:
in an embodiment of the present invention, a computer-readable storage medium is provided, which stores a computer program, which when executed by a processor implements the steps in the above-mentioned deep learning cardiac CT image segmentation method embodiment, for example, steps S101 to S103 shown in fig. 1. Alternatively, the computer program may be adapted to perform the functions of the units of the above-described device embodiments, such as the functions of the units 21 to 23 shown in fig. 2, when executed by the processor.
In the embodiment of the invention, when a heart CT image segmentation request is received, a heart CT image input by a user is acquired, the acquired heart CT image is preprocessed to acquire a corresponding preprocessed image, and a pre-trained V-Net model is used for carrying out image segmentation on a preset heart tissue region in the preprocessed image to acquire a heart tissue region segmentation image corresponding to the heart CT image, so that the accuracy of image segmentation on the heart CT image is improved, a high-precision segmentation image is acquired, and the safety degree of an operation is improved.
The computer readable storage medium of the embodiments of the present invention may include any entity or device capable of carrying computer program code, a recording medium, such as a ROM/RAM, a magnetic disk, an optical disk, a flash memory, or the like.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.
Claims (8)
1. A heart CT image segmentation method based on deep learning is characterized by comprising the following steps:
when a segmentation request of a cardiac CT image is received, acquiring the cardiac CT image input by a user;
preprocessing the acquired cardiac CT image to obtain a corresponding preprocessed image;
performing image segmentation on a preset heart tissue region in the preprocessed image through a pre-trained V-Net model to obtain a heart tissue region segmentation image corresponding to the heart CT image; wherein the preset cardiac tissue region comprises a left ventricle of the heart;
before the step of acquiring the cardiac CT image input by the user, the method further comprises:
performing image segmentation on the heart tissue area in a pre-acquired heart CT sample image through a preset image processing program to obtain a GT image corresponding to the heart tissue area in the heart CT sample image;
training the V-Net model according to the cardiac CT sample image and the GT image;
when the similarity degree of the heart tissue region segmentation image output by the trained V-Net model and the GT image meets a preset threshold value, finishing the training of the V-Net model;
and the output image and the input image of the V-Net model have the same dimension.
2. The method of claim 1, wherein the predetermined image processing procedure is ITK-SNAP.
3. The method of claim 1, wherein the V-Net model comprises one input layer, 4 compression layers, 4 decompression layers, and one output layer, wherein each layer in the V-Net model has a convolution kernel size of 5, wherein the convolution step size of the input layer is 1, and wherein the convolution step size of the compression layers, the decompression layers, and the output layer is 2.
4. A cardiac CT image segmentation apparatus based on deep learning, the apparatus comprising:
the CT image acquisition unit is used for acquiring a cardiac CT image input by a user when a segmentation request of the cardiac CT image is received;
the image preprocessing unit is used for preprocessing the acquired cardiac CT image to acquire a corresponding preprocessed image; and
the image segmentation unit is used for carrying out image segmentation on a preset heart tissue region in the preprocessed image through a pre-trained V-Net model to obtain a heart tissue region segmentation image corresponding to the heart CT image; wherein the preset cardiac tissue region comprises a left ventricle of the heart;
the GT image acquisition unit is used for carrying out image segmentation on the heart tissue area in a heart CT sample image acquired in advance through a preset image processing program to obtain a GT image corresponding to the heart tissue area in the heart CT sample image;
the model training unit is used for training the V-Net model according to the heart CT sample image and the GT image; and
and the training ending unit is used for ending the training of the V-Net model when the similarity degree of the heart tissue region segmentation image output by the V-Net model after training and the GT image meets a preset threshold value.
5. The apparatus of claim 4, wherein the predetermined image processing program is ITK-SNAP.
6. The apparatus of claim 4, wherein the V-Net model comprises an input layer, 4 compression layers, 4 decompression layers, and an output layer, wherein each layer in the V-Net model has a convolution kernel size of 5, wherein the convolution step size of the input layer is 1, and wherein the convolution step size of the compression layers, the decompression layers, and the output layer is 2.
7. A computing device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the steps of the method according to any of claims 1 to 3 when executing the computer program.
8. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 3.
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 CN109035284A (en) | 2018-12-18 |
| CN109035284B true 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) |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| 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 |
| US11436720B2 (en) | 2018-12-28 | 2022-09-06 | Shanghai United Imaging Intelligence Co., Ltd. | Systems and methods for generating image metric |
| CN109949309B (en) * | 2019-03-18 | 2022-02-11 | 安徽紫薇帝星数字科技有限公司 | Liver CT image segmentation 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 |
| 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 |
| 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 |
| 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 |
| 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 |
| 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 |
| 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 (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7450746B2 (en) * | 2002-06-07 | 2008-11-11 | Verathon Inc. | System and method for cardiac imaging |
| 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 (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
|---|
| Segmentation of anatomical structures in cardiac CTA using multi-label V-Net;Tang H等;《Medical Imaging 2018: Image Processing. International Society for Optics and Photonics》;20180302;全文 * |
| V-net: Fully convolutional neural networks for volumetric medical image segmentation;Milletari F等;《2016 fourth international conference on 3D vision (3DV). IEEE》;20161231;第一、三章、图2-3 * |
| Volumetric multimodality neural network for brain tumor segmentation;Castillo L S等;《13th international conference on medical information processing and analysis. International Society for Optics and Photonics, 2017》;20171231;全文 * |
| 基于改进C-V模型的肾脏CT图像分割方法;张品等;《光电子.激光》;20130315(第03期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109035284A (en) | 2018-12-18 |
| WO2020001086A1 (en) | 2020-01-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109035284B (en) | Heart CT image segmentation method, device, equipment and medium based on deep learning | |
| EP3559903B1 (en) | Machine learning of anatomical model parameters | |
| EP4002269A1 (en) | Systems and methods for image-based object modeling using multiple image acquisitions or reconstructions | |
| US10098602B2 (en) | Apparatus and method for processing a medical image of a body lumen | |
| EP2365471B1 (en) | Diagnosis assisting apparatus, coronary artery analyzing method and recording medium having a coronary artery analzying program stored therein | |
| KR101902883B1 (en) | A method for analyzing plaque in a computed tomography image and an apparatus thereof | |
| US20190209113A1 (en) | Method, Storage Medium, and System for Analyzing Image Sequences of Periodic Physiological Activities | |
| CN109961419B (en) | Correction information acquisition method for attenuation correction of PET activity distribution image | |
| EP3971830A1 (en) | Pneumonia sign segmentation method and apparatus, medium and electronic device | |
| CN109102509B (en) | Segmentation model training method and device and computer readable storage medium | |
| CN113989231A (en) | Method and device for determining kinetic parameters, computer equipment and storage medium | |
| CN110739050B (en) | Left ventricle full-parameter and confidence coefficient quantification method | |
| CN112529919A (en) | System and method for generating a bull's eye map generation of a subject's heart | |
| US20230222668A1 (en) | Image processing apparatus, image processing method, and recording medium | |
| CN116168099A (en) | Medical image reconstruction method and device and nonvolatile storage medium | |
| CN112529977B (en) | PET image reconstruction method and system | |
| Shaaf et al. | A Convolutional Neural Network Model to Segment Myocardial Infarction from MRI Images. | |
| US20220044454A1 (en) | Deep reinforcement learning for computer assisted reading and analysis | |
| WO2019146356A1 (en) | Image processing device, image processing, and program | |
| EP4312229A1 (en) | Information processing apparatus, information processing method, program, trained model, and learning model generation method | |
| WO2023131061A1 (en) | Systems and methods for positron emission computed tomography image reconstruction | |
| WO2023040918A1 (en) | Methods and systems for vascular analysis | |
| JP2023083660A (en) | Myocardial energy calculation method, myocardial energy calculation system, myocardial energy calculation device, and myocardial energy calculation program | |
| CN113096210A (en) | Image reconstruction method and device, electronic equipment and storage medium | |
| KR20230049589A (en) | Method, device, computing device and storage medium for determining blood flow rate |
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 |