CN114612404A - Blood vessel segmentation method, device, storage medium and electronic equipment - Google Patents
Blood vessel segmentation method, device, storage medium and electronic equipment Download PDFInfo
- Publication number
- CN114612404A CN114612404A CN202210207828.9A CN202210207828A CN114612404A CN 114612404 A CN114612404 A CN 114612404A CN 202210207828 A CN202210207828 A CN 202210207828A CN 114612404 A CN114612404 A CN 114612404A
- Authority
- CN
- China
- Prior art keywords
- dimensional
- blood vessel
- image
- cutting
- layer
- 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
- 210000004204 blood vessel Anatomy 0.000 title claims abstract description 133
- 230000011218 segmentation Effects 0.000 title claims abstract description 80
- 238000000034 method Methods 0.000 title claims abstract description 65
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 31
- 238000012549 training Methods 0.000 claims description 21
- 238000002372 labelling Methods 0.000 claims description 19
- 238000004590 computer program Methods 0.000 claims description 9
- 238000002759 z-score normalization Methods 0.000 claims description 8
- 230000035755 proliferation Effects 0.000 claims description 5
- 238000013519 translation Methods 0.000 claims description 5
- 230000008569 process Effects 0.000 description 13
- 238000010586 diagram Methods 0.000 description 8
- 238000013135 deep learning Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 208000006011 Stroke Diseases 0.000 description 2
- 238000002583 angiography Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000003709 image segmentation Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 208000014882 Carotid artery disease Diseases 0.000 description 1
- 206010008190 Cerebrovascular accident Diseases 0.000 description 1
- 238000005481 NMR spectroscopy Methods 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 210000001367 artery Anatomy 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 239000008280 blood Substances 0.000 description 1
- 210000004369 blood Anatomy 0.000 description 1
- 208000037876 carotid Atherosclerosis Diseases 0.000 description 1
- 210000001715 carotid artery Anatomy 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000002490 cerebral effect Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000003902 lesion Effects 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000004445 quantitative analysis Methods 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 208000023516 stroke disease Diseases 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Images
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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06T7/155—Segmentation; Edge detection involving morphological operators
-
- 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
- G06T2207/30101—Blood vessel; Artery; Vein; Vascular
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a blood vessel segmentation method, a blood vessel segmentation device, a storage medium and an electronic device, wherein the blood vessel segmentation method comprises the following steps: acquiring a three-dimensional blood vessel image; cutting the three-dimensional blood vessel image to obtain a plurality of three-dimensional cutting blocks with preset sizes; and respectively inputting each three-dimensional cutting block into a pre-trained convolutional neural network model to obtain a blood vessel segmentation result corresponding to each three-dimensional cutting block, and splicing the blood vessel segmentation results corresponding to the three-dimensional cutting blocks to obtain a blood vessel segmentation result of the three-dimensional blood vessel image. The invention can realize the full-automatic and accurate segmentation of the three-dimensional blood vessel image, and the image quality of the segmentation result is higher; the image features are fully utilized, and the method can be used for segmenting scenes of high-resolution images.
Description
Technical Field
The present invention relates to the field of image processing, and in particular, to a method and an apparatus for segmenting blood vessels, a storage medium, and an electronic device.
Background
According to the latest statistics of the world health organization, stroke has become the second leading cause of death in the world, wherein carotid atherosclerosis is an important cause of stroke disease. Accurately evaluating the blood vessel morphology of the neck artery and carrying out quantitative analysis and diagnosis, and is beneficial to the prevention of cerebral apoplexy.
Time of flight Magnetic resonance angiography (TOF MRA) is a nuclear Magnetic resonance angiography technique that generates a low signal in a stationary tissue but a high signal in flowing blood based on an inflow enhancement effect, has the advantages of safety, no radiation, high imaging speed, high contrast, high spatial resolution, large coverage and the like, and is an important clinical examination means for carotid artery lesions. Vessel segmentation and visualization based on three-dimensional TOF MRA images is key to describing vessel morphology. With the increase of the number of patients and the shortage of the number of professional physicians, the computer-aided magnetic resonance blood vessel image for blood vessel segmentation has become an important development direction.
With the continuous improvement of the traditional method and the development of artificial intelligence technology, some full-automatic vessel segmentation methods based on two-dimensional TOF MRA or three-dimensional TOF MRA images are proposed successively, but the methods are difficult to mine the image features of the TOF MRA and have unsatisfactory segmentation effect.
In the related technology, the full-automatic vessel segmentation method based on the two-dimensional TOF MRA or three-dimensional TOF MRA image is mainly divided into a traditional algorithm and a deep learning algorithm, wherein the traditional algorithm mainly comprises the steps of performing foreground and background segmentation by using a Dajin threshold value, performing vessel segmentation by fitting a mixed distribution model and automatically selecting seed points to perform region growth, and the deep learning algorithm mainly performs learning and modeling of a segmentation task based on an advanced U-Net or 3D U-Net model.
However, the prior art has the following disadvantages:
a. the traditional full-automatic segmentation method is low in accuracy and high in requirement on image quality.
b. The existing deep learning segmentation method does not fully utilize the image characteristics of TOF MRA, and the accuracy improvement is very limited.
c. The model has poor portability, is easily limited by computer hardware, and is difficult to be used for a segmentation scene of a high-resolution image.
Disclosure of Invention
In order to solve the above problems, embodiments of the present invention provide a blood vessel segmentation method, device, storage medium, and electronic device, which can implement full-automatic and accurate segmentation of a three-dimensional blood vessel image, and the segmentation result has high image quality; the image features are fully utilized, and the method can be used for segmenting scenes of high-resolution images.
In a first aspect, an embodiment of the present invention provides a blood vessel segmentation method, including:
acquiring a three-dimensional blood vessel image;
cutting the three-dimensional blood vessel image to obtain a plurality of three-dimensional cutting blocks with preset sizes;
and respectively inputting each three-dimensional cutting block into a pre-trained convolutional neural network model to obtain a blood vessel segmentation result corresponding to each three-dimensional cutting block, and splicing the blood vessel segmentation results corresponding to the three-dimensional cutting blocks to obtain a blood vessel segmentation result of the three-dimensional blood vessel image.
In some embodiments, the convolutional neural network model comprises a 3D U-Net model; the 3DU-Net model comprises four coding layers and four decoding layers, and a residual error structure is introduced into each coding layer and each decoding layer, so that the feature diagram of the previous layer is combined with the feature diagram after convolution of the layer and then is transmitted to the next layer.
In some embodiments, each layer of skip connection of the 3D U-Net model is added with a three-dimensional cut block down-sampled to the layer, so that the feature map output by the encoder of each layer and the features of the three-dimensional cut block down-sampled to the layer are combined and then sent to the decoder of the corresponding layer.
In some embodiments, the method further comprises:
acquiring an original three-dimensional blood vessel image and a corresponding labeled blood vessel labeling image;
cutting the original three-dimensional blood vessel image and the corresponding marked blood vessel marking image to obtain a plurality of three-dimensional cutting blocks with preset sizes;
and training a convolutional neural network model by taking the three-dimensional blocks of the original three-dimensional blood vessel image as input and the three-dimensional blocks of the labeled blood vessel labeling image as output.
In some embodiments, the cutting the original three-dimensional blood vessel image and the corresponding labeled blood vessel labeling image to obtain a plurality of three-dimensional cut blocks with preset sizes includes:
and respectively carrying out translation with different step lengths in three directions of the original three-dimensional blood vessel image and the corresponding marked blood vessel marked image in a data proliferation mode, and cutting out a plurality of three-dimensional cutting blocks with preset sizes.
In some embodiments, before training the convolutional neural network model with the three-dimensional slice of the original three-dimensional blood vessel image as input and the three-dimensional slice of the labeled blood vessel labeling image as output, the method further comprises:
each three-dimensional cut was subjected to Z-score normalization.
In some embodiments, the subjecting each three-dimensional slice to a Z-score normalization process comprises:
and subtracting the average value of all voxel values of the current three-dimensional block from each voxel value of each three-dimensional block, dividing the average value by the standard deviation of all voxel values of the current three-dimensional block, and replacing the obtained result with the current voxel value of the current three-dimensional block.
In a second aspect, an embodiment of the present invention provides a blood vessel segmentation apparatus, including:
the image acquisition module is used for acquiring a three-dimensional blood vessel image;
the image cutting module is used for cutting the three-dimensional blood vessel image to obtain a plurality of three-dimensional cutting blocks with preset sizes;
and the blood vessel segmentation module is used for respectively inputting each three-dimensional cutting block into a pre-trained convolutional neural network model to obtain a blood vessel segmentation result corresponding to each three-dimensional cutting block, and the blood vessel segmentation results corresponding to the three-dimensional cutting blocks are spliced to obtain a blood vessel segmentation result of the three-dimensional blood vessel image.
In a third aspect, an embodiment of the present invention provides a computer-readable storage medium, including: the computer-readable storage medium has stored thereon a computer program which, when executed by one or more processors, implements the vessel segmentation method according to the first aspect.
In a fourth aspect, an embodiment of the present invention provides an electronic device, including: comprising a memory and one or more processors, the memory having stored thereon a computer program which, when executed by the one or more processors, carries out the vessel segmentation method as set forth in the first aspect.
Compared with the prior art, one or more embodiments of the invention have at least the following beneficial effects:
according to the blood vessel segmentation method, the blood vessel segmentation device, the storage medium and the electronic equipment, due to the fact that training and prediction are based on three-dimensional blocks, for images with different sizes, the improved convolutional neural network can be trained and predicted, a segmentation mask image with the size consistent with that of an input image is output, and the problem of universality of the size of the image predicted by model training is solved; the convolutional neural network structure is effectively improved, and more image features are introduced, so that systematic and purposeful learning and prediction of image data are realized, and the accuracy of blood vessel segmentation is improved. The three-dimensional blood vessel image is learned and modeled through the improved convolutional neural network, automatic and correct blood vessel segmentation of new image data is achieved, assistance is provided for a doctor to obtain morphological characteristics of the blood vessel, and the working efficiency of the doctor can be greatly improved. The method can be popularized to the 3D TOF MRA blood vessel segmentation process of other parts.
Drawings
To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are required to be used in the embodiments will be briefly described below, and it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope.
FIG. 1 is a flow chart of a vessel segmentation method provided by an embodiment of the invention;
FIG. 2 is a schematic structural diagram of a 3D U-Net model provided by an embodiment of the invention;
fig. 3 is a block diagram of a blood vessel segmentation apparatus according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present invention without making any creative effort, shall fall within the protection scope of the present invention.
Example one
Fig. 1 is a flowchart of a blood vessel segmentation method according to an embodiment of the present invention, and as shown in fig. 1, the blood vessel segmentation method according to the embodiment at least includes steps S101 to S103:
and step S101, acquiring a three-dimensional blood vessel image.
In practice, the three-dimensional vessel image may be, but is not limited to, a three-dimensional TOF MRA image.
And S102, cutting the three-dimensional blood vessel image to obtain a plurality of three-dimensional cutting blocks with preset sizes.
Due to the limitation of computer hardware (such as the memory size of a display card), three-dimensional blocks with the size meeting the operation requirement of the computer hardware are cut out from the original three-dimensional TOF MRA image, the size of the selected blocks is as large as possible, and the preset size can be determined according to the actual situation. And after cutting, predicting the segmentation result of each three-dimensional cutting block by using a convolutional neural network model, and splicing the predicted segmentation results to obtain a complete segmentation mask image as a blood vessel segmentation result.
Step S103, inputting each three-dimensional block into a pre-trained convolutional neural network model respectively to obtain a blood vessel segmentation result corresponding to each three-dimensional block, and splicing the blood vessel segmentation results corresponding to the three-dimensional blocks to obtain a blood vessel segmentation result of a three-dimensional blood vessel image.
And after each three-dimensional block cut out from the three-dimensional TOF MRA image is input into a convolutional neural network model, continuously extracting the characteristics of different levels through convolution and pooling operations so as to complete final prediction. In the convolutional neural network model, the former convolutional layer will extract the preliminary local information of the (three-dimensional block) image, and the latter convolutional layer can extract the global features of higher level, which can be combined to distinguish the boundary of the blood vessel from other tissues and obtain the continuity between the blood vessels.
In the method of the embodiment, because the training and prediction based on the three-dimensional block are adopted, the convolutional neural network can train and predict images with different sizes, and output the segmentation mask image with the size consistent with that of the input image, thereby solving the problem of the universality of the size of the image predicted by model training.
The convolutional neural network model described above includes a 3D U-Net model, as shown in fig. 2.
The 3D U-Net model comprises four coding layers and four decoding layers, and each coding layer and each decoding layer introduce a residual error structure so as to combine the feature map of the previous layer with the feature map after convolution of the layer and then transmit the feature map to the next layer. Each convolution operation employs a Relu activation function.
In the embodiment, a coding layer and a decoding layer are added on the original 3D U-Net structure, and the coding layer and the decoding layer are improved to be four coding layers and four decoding layers to deal with the large-size input image, so that the down-sampling rate of the network is 2 to the power of 4 and is equal to 16 times, and deeper features can be acquired in the large-size image. The residual error structure is added in the convolution operation of each layer, the feature diagram of the previous layer can be combined with the feature diagram of the layer after convolution, and then the combined feature diagram is transmitted to the next layer.
In some embodiments, each layer of skip connection of the 3D U-Net model is added with a three-dimensional cut block down-sampled to the layer, so that the feature map output by the encoder of each layer and the features of the three-dimensional cut block down-sampled to the layer are combined and then sent to the decoder of the corresponding layer.
And after 1-time, 2-time, 4-time and 8-time down-sampling is respectively carried out on the input TOF MRA image blocks, the TOF MRA image blocks are combined with the feature map output by each layer of encoder, and the self-carrying features of the TOF MRA image blocks and the features extracted by the encoder are transmitted to a decoder of a corresponding hierarchy level together for training and learning through skipping connection. Due to the adoption of the structure, the network can receive more TOF MRA image information, better learn the image characteristics of the original image and better improve the generalization capability of the network to TOF MRA image segmentation tasks.
Further, the method further comprises step S201 to step S203:
step S201, an original three-dimensional blood vessel image and a corresponding labeled blood vessel labeling image are obtained.
The convolutional neural network model in the embodiment belongs to a deep learning type with supervised learning, an original TOF MRA image and a corresponding professional blood vessel labeling image need to be provided in the training process of the model, and the convolutional neural network establishes high-level connection between the original image and the corresponding labeling image through a large number of sample learning. In the model training and learning process, the labeled blood vessel labeling image is input into the convolutional neural network for model training, and the model obtained after model training can be used for predicting a new sample in the future, so that a blood vessel segmentation result of a TOF MRA image (such as a neck three-dimensional TOF MRA image) is obtained.
Step S202, cutting the original three-dimensional blood vessel image and the corresponding marked blood vessel marking image to obtain a plurality of three-dimensional blocks with preset sizes.
Due to the limitation of computer hardware (such as the size of a memory of a display card), three-dimensional blocks with the size meeting the operation requirement of the computer hardware are cut out from the original three-dimensional TOF MRA image, the size of the selected blocks is as large as possible, and the preset size can be determined according to actual conditions.
In some embodiments, the clipping the original three-dimensional blood vessel image and the corresponding labeled blood vessel labeling image to obtain a plurality of three-dimensional cut blocks with preset sizes includes:
and respectively carrying out translation with different step lengths in three directions of the original three-dimensional blood vessel image and the corresponding marked blood vessel marking image in a data proliferation mode, and cutting out a plurality of three-dimensional cutting blocks with preset sizes.
By carrying out data proliferation on the original three-dimensional TOF MRA image and the blood vessel labeling image, the generalization capability of the model is improved, and the problem of overfitting of the model in the training process is avoided.
It should be understood that, the translation with different step lengths is randomly performed in three directions of the three-dimensional blood vessel image in a data multiplication manner, and a plurality of three-dimensional blocks with preset sizes are cut out, so that the method is suitable for the original three-dimensional blood vessel image and the labeled blood vessel labeling image.
Step S203, taking the three-dimensional blocks of the original three-dimensional blood vessel image as input and the three-dimensional blocks of the labeled blood vessel labeling image as output, and training a convolutional neural network model.
As an end-to-end training reasoning method, the convolutional neural network can automatically extract and process features in a model, and no manual operation is needed in the operation process. Therefore, only the original TOF MRA image needs to be input into the convolutional neural network, so that the learning modeling can be automatically performed, and the segmentation result is output. The complex convolution module and post-processing steps are avoided, and the end-to-end segmentation task is realized.
In some cases, before training the convolutional neural network model with the three-dimensional slice of the original three-dimensional blood vessel image as input and the three-dimensional slice of the labeled blood vessel labeling image as output, the method further comprises: each three-dimensional cut was subjected to Z-score normalization.
Further, each three-dimensional cut is subjected to a Z-score normalization process comprising:
and subtracting the average value of all voxel values of the current three-dimensional block from each voxel value of each three-dimensional block, dividing the average value by the standard deviation of all voxel values of the current three-dimensional block, and replacing the obtained result with the current voxel value of the current three-dimensional block.
Compared with the existing 3D U-Net, the 3D U-Net model in the embodiment is improved in at least the following aspects:
(1) aiming at the high-resolution large-size input image, a layer of encoder and decoder is added, the network structure is deepened, and more advanced features are extracted.
(2) And residual error structures are introduced into each layer of coder and decoder, so that the problem of network degradation in the network deepening process is avoided.
(3) By adding the original image which is down-sampled to the size same as that of the hierarchy feature map in each layer of skip connection, the added original image can play a gating effect of regional response enhancement, information of irrelevant regions is suppressed, and a target region is more highlighted, so that the feature learning capability of a network on TOF MRA images is improved, and the generalization capability of the network on TOF MRA image segmentation tasks is improved.
The convolution neural network model obtained by training in the steps can realize full-automatic and accurate segmentation of the three-dimensional blood vessel image, and the segmentation result has high image quality; the image characteristics of TOF MRA are fully utilized, the accuracy rate is obviously improved, and in addition, the model has better transportability and is not limited by computer hardware, so that the model can be used for segmenting scenes of high-resolution images.
Example two
Fig. 3 is a block diagram of a blood vessel segmentation apparatus according to the present embodiment, and the blood vessel segmentation apparatus according to the present embodiment, as shown in fig. 3, includes:
an image acquisition module 301, configured to acquire a three-dimensional blood vessel image;
the image cutting module 302 is configured to cut a three-dimensional blood vessel image to obtain a plurality of three-dimensional cut blocks with preset sizes;
and the blood vessel segmentation module 303 is configured to input each three-dimensional slice into a pre-trained convolutional neural network model, so as to obtain a blood vessel segmentation result corresponding to each three-dimensional slice, and the blood vessel segmentation results corresponding to the three-dimensional slices are pieced together to obtain a blood vessel segmentation result of the three-dimensional blood vessel image.
In some embodiments, the convolutional neural network model comprises a 3D U-Net model. The 3D U-Net model comprises four coding layers and four decoding layers, and each coding layer and each decoding layer introduce a residual error structure so as to combine the feature map of the previous layer with the feature map after convolution of the layer and then transmit the feature map to the next layer.
In some embodiments, each layer of skip connection of the 3D U-Net model is added with a three-dimensional cut block down-sampled to the layer, so that the feature map output by the encoder of each layer and the features of the three-dimensional cut block down-sampled to the layer are combined and then sent to the decoder of the corresponding layer.
Further, the apparatus may further include:
the model training module is used for acquiring an original three-dimensional blood vessel image and a corresponding labeled blood vessel labeling image; cutting an original three-dimensional blood vessel image and a corresponding marked blood vessel marking image to obtain a plurality of three-dimensional blocks with preset sizes; and training a convolutional neural network model by taking the three-dimensional blocks of the original three-dimensional blood vessel image as input and the three-dimensional blocks of the labeled blood vessel labeling image as output.
Cutting an original three-dimensional blood vessel image and a corresponding marked blood vessel marking image to obtain a plurality of three-dimensional cutting blocks with preset sizes, wherein the cutting method comprises the following steps: and respectively carrying out translation with different step lengths in three directions of the original three-dimensional blood vessel image and the corresponding marked blood vessel marking image in a data proliferation mode, and cutting out a plurality of three-dimensional cutting blocks with preset sizes.
In some cases, before training the convolutional neural network model with the three-dimensional slice of the original three-dimensional blood vessel image as input and the three-dimensional slice of the labeled blood vessel labeling image as output, the method further comprises: each three-dimensional cut was subjected to Z-score normalization.
Further, each three-dimensional cut is subjected to a Z-score normalization process comprising:
and subtracting the average value of all voxel values of the current three-dimensional block from each voxel value of each three-dimensional block, dividing the average value by the standard deviation of all voxel values of the current three-dimensional block, and replacing the obtained result with the current voxel value of the current three-dimensional block.
It should be understood that the apparatus of the present embodiment provides all of the benefits of the method embodiments.
Those skilled in the art will appreciate that the modules or steps described above can be implemented using a general purpose computing device, that they can be centralized on a single computing device or distributed across a network of computing devices, and that they can alternatively be implemented using program code executable by a computing device, such that the program code is stored in a memory device and executed by a computing device, and the program code is then separately fabricated into various integrated circuit modules, or multiple modules or steps are fabricated into a single integrated circuit module. The present invention is not limited to any specific combination of hardware and software.
EXAMPLE III
The embodiment of the invention provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by one or more processors, the blood vessel segmentation method of the first embodiment is implemented.
In this embodiment, the storage medium may be implemented by any type of volatile or nonvolatile storage device or combination thereof, such as a Static Random Access Memory (SRAM), an Electrically Erasable Programmable Read-Only Memory (EEPROM), an Erasable Programmable Read-Only Memory (EPROM), a Programmable Read-Only Memory (PROM), a Read-Only Memory (ROM), a magnetic Memory, a flash Memory, a magnetic disk, or an optical disk.
Example four
The embodiment provides an electronic device, which includes a memory and one or more processors, where the memory stores a computer program, and the computer program is executed by the one or more processors to implement the blood vessel segmentation method of the first embodiment.
In practical application, the electronic device may be a mobile phone, a tablet computer, or other terminal device. In this embodiment, the Processor may be an Application Specific Integrated Circuit (ASIC), a Digital Signal Processor (DSP), a Digital Signal Processing Device (DSPD), a Programmable Logic Device (PLD), a Field Programmable Gate Array (FPGA), a controller, a microcontroller, a microprocessor, or other electronic components, and is configured to perform the method in the above embodiments. The method implemented when the computer program running on the processor is executed may refer to the specific embodiment of the method provided in the foregoing embodiment of the present invention, and details thereof are not described herein.
In the embodiments provided in the present invention, it should be understood that the disclosed system and method can be implemented in other ways. The system and method embodiments described above are merely illustrative.
It should be noted that, in this document, the terms "first", "second", and the like in the description and claims of the present application and in the drawings described above are used for distinguishing similar objects, and are not necessarily used for describing a particular order or sequence. The terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
Although the embodiments of the present invention have been described above, the above description is only for the purpose of understanding the present invention, and is not intended to limit the present invention. It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (10)
1. A method of vessel segmentation, comprising:
acquiring a three-dimensional blood vessel image;
cutting the three-dimensional blood vessel image to obtain a plurality of three-dimensional cutting blocks with preset sizes;
and respectively inputting each three-dimensional cutting block into a pre-trained convolutional neural network model to obtain a blood vessel segmentation result corresponding to each three-dimensional cutting block, and splicing the blood vessel segmentation results corresponding to the three-dimensional cutting blocks to obtain a blood vessel segmentation result of the three-dimensional blood vessel image.
2. The vessel segmentation method according to claim 1, wherein the convolutional neural network model comprises a 3D U-Net model; the 3D U-Net model comprises four coding layers and four decoding layers, and a residual error structure is introduced into each coding layer and each decoding layer, so that the feature map of the previous layer is combined with the feature map after convolution of the layer and then is transmitted to the next layer.
3. The vessel segmentation method according to claim 2, wherein a three-dimensional slice down-sampled to the current layer is added to each layer skip connection of the 3D U-Net model, so that the feature map output by the encoder of each layer and the features of the three-dimensional slice down-sampled to the current layer are combined and then sent to the decoder of the corresponding layer.
4. The blood vessel segmentation method according to claim 1, further comprising:
acquiring an original three-dimensional blood vessel image and a corresponding labeled blood vessel labeling image;
cutting the original three-dimensional blood vessel image and the corresponding marked blood vessel marking image to obtain a plurality of three-dimensional cutting blocks with preset sizes;
and training a convolutional neural network model by taking the three-dimensional blocks of the original three-dimensional blood vessel image as input and the three-dimensional blocks of the labeled blood vessel labeling image as output.
5. The method of claim 4, wherein the cutting the original three-dimensional blood vessel image and the labeled blood vessel image to obtain a plurality of three-dimensional blocks with preset sizes comprises:
and respectively carrying out translation with different step lengths in three directions of the original three-dimensional blood vessel image and the corresponding marked blood vessel marked image in a data proliferation mode, and cutting out a plurality of three-dimensional cutting blocks with preset sizes.
6. The method of claim 4, further comprising, before training the convolutional neural network model using the three-dimensional slice of the original three-dimensional blood vessel image as an input and the three-dimensional slice of the labeled blood vessel labeling image as an output:
each three-dimensional cut was subjected to Z-score normalization.
7. The method of claim 6, wherein the Z-score normalization of each three-dimensional slice comprises:
and subtracting the average value of all voxel values of the current three-dimensional block from each voxel value of each three-dimensional block, dividing the average value by the standard deviation of all voxel values of the current three-dimensional block, and replacing the obtained result with the current voxel value of the current three-dimensional block.
8. A vessel segmentation device, comprising:
the image acquisition module is used for acquiring a three-dimensional blood vessel image;
the image cutting module is used for cutting the three-dimensional blood vessel image to obtain a plurality of three-dimensional cutting blocks with preset sizes;
and the blood vessel segmentation module is used for respectively inputting each three-dimensional cutting block into a pre-trained convolutional neural network model to obtain a blood vessel segmentation result corresponding to each three-dimensional cutting block, and the blood vessel segmentation results corresponding to the three-dimensional cutting blocks are spliced to obtain a blood vessel segmentation result of the three-dimensional blood vessel image.
9. A computer-readable storage medium, comprising: the computer-readable storage medium has stored thereon a computer program which, when executed by one or more processors, implements the vessel segmentation method as claimed in any one of claims 1 to 7.
10. An electronic device, comprising: comprising a memory and one or more processors, the memory having stored thereon a computer program which, when executed by the one or more processors, implements a vessel segmentation method as claimed in any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210207828.9A CN114612404A (en) | 2022-03-04 | 2022-03-04 | Blood vessel segmentation method, device, storage medium and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210207828.9A CN114612404A (en) | 2022-03-04 | 2022-03-04 | Blood vessel segmentation method, device, storage medium and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114612404A true CN114612404A (en) | 2022-06-10 |
Family
ID=81861788
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210207828.9A Pending CN114612404A (en) | 2022-03-04 | 2022-03-04 | Blood vessel segmentation method, device, storage medium and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114612404A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115457024A (en) * | 2022-10-10 | 2022-12-09 | 水木未来(杭州)科技有限公司 | Method and device for processing cryoelectron microscope image, electronic equipment and storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108629784A (en) * | 2018-05-08 | 2018-10-09 | 上海嘉奥信息科技发展有限公司 | A kind of CT image intracranial vessel dividing methods and system based on deep learning |
| CN109816666A (en) * | 2019-01-04 | 2019-05-28 | 三峡大学 | Symmetrical full convolutional neural networks model building method, eye fundus image blood vessel segmentation method, apparatus, computer equipment and storage medium |
| US20200074271A1 (en) * | 2018-08-29 | 2020-03-05 | Arizona Board Of Regents On Behalf Of Arizona State University | Systems, methods, and apparatuses for implementing a multi-resolution neural network for use with imaging intensive applications including medical imaging |
| US20210142470A1 (en) * | 2019-11-12 | 2021-05-13 | International Intelligent Informatics Solution Laboratory LLC | System and method for identification of pulmonary arteries and veins depicted on chest ct scans |
| CN112950643A (en) * | 2021-02-26 | 2021-06-11 | 东北大学 | New coronary pneumonia focus segmentation method based on feature fusion deep supervision U-Net |
| CN113012166A (en) * | 2021-03-19 | 2021-06-22 | 北京安德医智科技有限公司 | Intracranial aneurysm segmentation method and device, electronic device, and storage medium |
-
2022
- 2022-03-04 CN CN202210207828.9A patent/CN114612404A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108629784A (en) * | 2018-05-08 | 2018-10-09 | 上海嘉奥信息科技发展有限公司 | A kind of CT image intracranial vessel dividing methods and system based on deep learning |
| US20200074271A1 (en) * | 2018-08-29 | 2020-03-05 | Arizona Board Of Regents On Behalf Of Arizona State University | Systems, methods, and apparatuses for implementing a multi-resolution neural network for use with imaging intensive applications including medical imaging |
| CN109816666A (en) * | 2019-01-04 | 2019-05-28 | 三峡大学 | Symmetrical full convolutional neural networks model building method, eye fundus image blood vessel segmentation method, apparatus, computer equipment and storage medium |
| US20210142470A1 (en) * | 2019-11-12 | 2021-05-13 | International Intelligent Informatics Solution Laboratory LLC | System and method for identification of pulmonary arteries and veins depicted on chest ct scans |
| CN112950643A (en) * | 2021-02-26 | 2021-06-11 | 东北大学 | New coronary pneumonia focus segmentation method based on feature fusion deep supervision U-Net |
| CN113012166A (en) * | 2021-03-19 | 2021-06-22 | 北京安德医智科技有限公司 | Intracranial aneurysm segmentation method and device, electronic device, and storage medium |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115457024A (en) * | 2022-10-10 | 2022-12-09 | 水木未来(杭州)科技有限公司 | Method and device for processing cryoelectron microscope image, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11170504B2 (en) | Method and system for intracerebral hemorrhage detection and segmentation based on a multi-task fully convolutional network | |
| CN109872306B (en) | Medical image segmentation method, device and storage medium | |
| CN111161279B (en) | Medical image segmentation method, device and server | |
| CN111028242A (en) | Automatic tumor segmentation system and method and electronic equipment | |
| CN109978037A (en) | Image processing method, model training method, device and storage medium | |
| CN105574859A (en) | Liver tumor segmentation method and device based on CT (Computed Tomography) image | |
| CN110310287A (en) | It is neural network based to jeopardize the automatic delineation method of organ, equipment and storage medium | |
| Nguyen et al. | Evaluating transformer-based semantic segmentation networks for pathological image segmentation | |
| CN110047068A (en) | MRI brain tumor dividing method and system based on pyramid scene analysis network | |
| CN110866921A (en) | Weakly supervised vertebral body segmentation method and system based on self-training and slice propagation | |
| Wazir et al. | HistoSeg: Quick attention with multi-loss function for multi-structure segmentation in digital histology images | |
| EP3973508A1 (en) | Sampling latent variables to generate multiple segmentations of an image | |
| Wang et al. | Left atrial appendage segmentation based on ranking 2-D segmentation proposals | |
| CN107194933A (en) | With reference to convolutional neural networks and the brain tumor dividing method and device of fuzzy reasoning | |
| CN114612404A (en) | Blood vessel segmentation method, device, storage medium and electronic equipment | |
| Zhang et al. | Autonomous localization and segmentation for body composition quantization on abdominal CT | |
| Muñoz-Benavent et al. | Impact evaluation of deep learning on image segmentation for automatic bluefin tuna sizing | |
| CN117437423A (en) | Weak supervision medical image segmentation method and device based on SAM collaborative learning and cross-layer feature aggregation enhancement | |
| CN111860465A (en) | Remote sensing image extraction method, device, equipment and storage medium based on super pixels | |
| CN116486071A (en) | Image blocking feature extraction method, device and storage medium | |
| CN115908299A (en) | Medical image-based life cycle prediction method, device, equipment and medium | |
| CN114972382A (en) | Brain tumor segmentation algorithm based on lightweight UNet + + network | |
| Ammari et al. | Deep-active-learning approach towards accurate right ventricular segmentation using a two-level uncertainty estimation | |
| CN114913145B (en) | Image segmentation method and device, electronic equipment and storage medium | |
| EP4343786A1 (en) | Building a machine-learning model to predict sematic context information for contrast-enhanced medical imaging measurements |
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 |