WO2022105735A1 - 冠状动脉分割方法及其装置、电子设备及计算机可读存储介质 - Google Patents
冠状动脉分割方法及其装置、电子设备及计算机可读存储介质 Download PDFInfo
- Publication number
- WO2022105735A1 WO2022105735A1 PCT/CN2021/130861 CN2021130861W WO2022105735A1 WO 2022105735 A1 WO2022105735 A1 WO 2022105735A1 CN 2021130861 W CN2021130861 W CN 2021130861W WO 2022105735 A1 WO2022105735 A1 WO 2022105735A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- data
- coronary artery
- segmentation
- detected
- coronary
- Prior art date
Links
- 210000004351 coronary vessel Anatomy 0.000 title claims abstract description 456
- 230000011218 segmentation Effects 0.000 title claims abstract description 340
- 238000000034 method Methods 0.000 title claims abstract description 81
- 238000001514 detection method Methods 0.000 claims abstract description 147
- 210000004204 blood vessel Anatomy 0.000 claims description 104
- 238000003062 neural network model Methods 0.000 claims description 92
- 238000012549 training Methods 0.000 claims description 44
- 238000013528 artificial neural network Methods 0.000 claims description 22
- 238000004590 computer program Methods 0.000 claims description 15
- 238000012545 processing Methods 0.000 claims description 14
- 238000000605 extraction Methods 0.000 claims description 8
- 238000013075 data extraction Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 12
- 210000003462 vein Anatomy 0.000 description 11
- 230000000747 cardiac effect Effects 0.000 description 10
- 238000013135 deep learning Methods 0.000 description 8
- 230000008569 process Effects 0.000 description 5
- 210000000709 aorta Anatomy 0.000 description 4
- 238000010968 computed tomography angiography Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 238000002591 computed tomography Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 208000029078 coronary artery disease Diseases 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 230000002792 vascular Effects 0.000 description 2
- 208000024172 Cardiovascular disease Diseases 0.000 description 1
- 238000007792 addition Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000017531 blood circulation Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 230000003902 lesion Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Definitions
- the present application relates to the technical field of deep learning, and in particular, to a coronary artery segmentation method, a coronary artery segmentation device, an electronic device, and a computer-readable storage medium.
- Coronary artery segmentation is crucial for coronary heart disease screening.
- the existing coronary artery segmentation methods are difficult to effectively distinguish coronary arteries and vein false positives, and are difficult to segment when there are motion artifacts or coronary plaques. Problems such as small coronary branches.
- the embodiments of the present application provide a coronary artery segmentation method, a coronary artery segmentation device, an electronic device, and a computer-readable storage medium, so as to solve the problem of difficulty in effectively distinguishing coronary arteries and veins in the coronary artery segmentation methods in the prior art.
- Technical issues such as false positives and difficulty in segmenting small coronary branches in the presence of motion artifacts or coronary plaque.
- an embodiment of the present application provides a method for segmenting coronary arteries, comprising: acquiring a region to be detected and regional coronary apocalypse data, and the regional coronary apocalypse data is used to provide enlightenment when segmenting the region to be detected ; Input the region to be detected with regional coronary apocalypse data into the pre-trained coronary artery segmentation model to obtain regional coronary artery segmentation data; obtain the comparison result between the regional coronary artery segmentation data and the regional coronary apocalypse data; use the comparison The result replaces the to-be-detected area in the to-be-detected image, and obtains the replaced to-be-detected image to complete the detection step; repeats the detection step, wherein, for the multiple detection steps corresponding to the to-be-detected image, after the replacement in the previous detection step Select the area to be detected of the next detection step in the image to be
- the region to be detected is the region of origin of the coronary artery; wherein, acquiring the region to be detected and the regional coronary apocalypse data includes: acquiring segmentation data of the origin of the coronary artery based on the image to be detected; a skeleton; select the first seed point on the first skeleton according to the first preset step size; push the first seed into the seed point stack to obtain the seed point at the top of the first stack; take the center of the seed point at the top of the first stack as the In the center, on the to-be-detected image, a first region of a first preset size is selected as the to-be-detected region; and the coronary artery origin segmentation data in the first region is used as regional coronary artery enlightenment data.
- the region to be detected comes from the replaced image to be detected in the previous detection step of the current detection step; wherein, acquiring the region to be detected and regional coronary apocalypse data includes: the ratio of the ratio in the previous detection step Obtain the newly added coronary artery segmentation data from the result; extract the second skeleton in the newly added coronary artery segmentation data; select the second seed point on the second skeleton according to the second preset step size; press the second seed point into the seed In the point stack, obtain the second stack top seed point; take the center of the second stack top seed point as the center, select the second area of the second preset size on the replaced image to be detected in the previous detection step As the region to be detected, the second region covers part of the comparison result in the previous detection step; and the comparison result in the previous detection step in the second region is used as regional coronary apocalypse data.
- obtaining the comparison result between the regional coronary artery segmentation data and the regional coronary apocalypse data includes: when the regional coronary artery segmentation data has more segmented regions than the regional coronary artery enlightenment data, selecting the regional coronary artery segmentation data is the comparison result; or when the regional coronary artery segmentation data does not segment more segmentation regions than the regional coronary apocalypse data, select the regional coronary artery apocalypse data as the comparison result.
- a method for training a coronary artery segmentation model includes: acquiring a sample, where the sample includes blood vessel data of a region to be identified and coronary artery identification data; adding interference data to the blood vessel data of the region to be identified, and obtaining expanded data of the blood vessel to be identified; Coronary artery identification data obtains coronary apocalypse data, and the coronary artery apocalypse data is used to provide an accurate origin point for the neural network model when identifying the blood vessel data in the to-be-identified area to provide enlightenment; Data and samples of coronary artery identification data are input into the neural network model, and the neural network model is trained so that the neural network model can output coronary artery identification data based on the coronary artery apocalypse data on the sample; wherein, the coronary artery is obtained based on the coronary artery identification data.
- the enlightenment data includes: selecting a part from the coronary artery identification data as the coronary artery
- adding interference data to the blood vessel data in the to-be-identified area includes: adding noise data to the blood vessel data in the to-be-identified area, where the noise data includes plaque data and artifact data; Add false positive data to the identified area vessel data.
- inputting the samples with the vessel expansion data to be identified, the coronary artery apocalypse data and the coronary artery identification data into the neural network model, and training the neural network model includes: adding the samples with the vessel expansion data to be identified and the coronary artery enlightenment data
- the samples of the neural network model are trained to obtain coronary artery segmentation data; loss results are obtained based on the coronary artery segmentation data and coronary artery identification data; and the parameters of the neural network model are adjusted based on the loss results.
- acquiring a sample includes: acquiring an original image, where the original image contains the identification of the blood vessel and coronary artery to be identified; and performing sliding window processing on the original image to acquire a plurality of samples.
- an embodiment of the present application provides a coronary artery segmentation device, comprising: a segmentation acquisition module configured to acquire a region to be detected and regional coronary apocalypse data, and the regional coronary apocalypse data is used to detect a region to be detected Provide enlightenment when the region is segmented; the segmentation module is configured to input the region to be detected with regional coronary apocalypse data into the pre-trained coronary artery segmentation model to obtain regional coronary artery segmentation data; the comparison and judgment module is configured to obtain the region The comparison result of the coronary apocalypse data and the regional coronary artery segmentation data; the replacement module is configured to replace the to-be-detected area in the to-be-detected image with the comparison result, and obtain the replaced to-be-detected image to complete the detection step; the output module, It is configured to repeat the detection step, wherein, for multiple detection steps corresponding
- the region to be detected is the region of origin of the coronary artery; wherein, the segmentation and acquisition module includes: a sub-module of origin segmentation, configured to acquire segmentation data of the origin of the coronary artery based on the image to be detected; the first extraction sub-module, configured to The first skeleton is extracted from the coronary artery origin segmentation data; the first seed point selection sub-module is configured to select the first seed point on the first skeleton according to the first preset step size; the first stack top seed point acquisition sub-module is configured In order to push the first seed into the seed point stack, obtain the seed point at the top of the first stack; the first area acquisition sub-module is configured to take the center of the seed point at the top of the first stack as the center, and select the first seed point on the image to be detected.
- the segmentation and acquisition module includes: a sub-module of origin segmentation, configured to acquire segmentation data of the origin of the coronary artery based on the image to be detected; the first extraction sub-modul
- a first region with a preset size is used as the region to be detected; and a first enlightenment acquisition sub-module is configured to use the segmented data of the origin of coronary arteries in the first area as regional coronary artery enlightenment data.
- the to-be-detected area comes from the replaced to-be-detected image in the previous detection step; wherein, the segmentation acquisition module further includes: a newly added acquisition sub-module, configured as the comparison result in the previous detection step
- the second extraction sub-module is configured to extract the second skeleton in the newly-added coronary artery segmentation data extraction;
- the second seed point selection sub-module is configured to follow the second preset on the second skeleton
- the step selects the second seed point;
- the second stack top seed point obtains a submodule, which is configured to push the second seed point into the seed point stack to obtain the second seed point at the top of the stack;
- the second area obtains the submodule, which is configured as follows:
- the center of the seed point at the top of the second stack is the center, and on the replaced image to be detected in the previous detection step, a second area of the second preset size is selected as the area to be detected, and the second area covers part of the previous detection area.
- the comparison judging module is further configured to select the regional coronary artery segmentation data as the comparison result when the regional coronary artery segmentation data divides more segmented regions than the regional coronary apocalypse data; or when the regional coronary artery segmentation data The data did not segment more segmented regions than the regional coronary apocalypse data, and the regional coronary apocalypse data was selected as the comparison result.
- the coronary artery segmentation device further includes: a sample acquisition module configured to acquire samples including blood vessel data in the region to be identified and coronary artery identification data; an expansion module configured to add interference data to the blood vessel data in the region to be identified , to obtain the expansion data of the blood vessel to be identified; the enlightenment acquisition module is configured to obtain the coronary apocalypse data based on the coronary artery identification data, and the coronary apocalypse data is used to provide an accurate origin point for the neural network model when identifying the blood vessel data in the area to be identified.
- obtaining coronary artery enlightenment data based on coronary artery identification data includes: selecting a part from coronary artery identification data as coronary artery enlightenment data; identification module, configured to be identified with vessel expansion data, coronary artery enlightenment data and The samples of coronary artery identification data are input into the neural network model, and the neural network model is trained so that the neural network model can output coronary artery identification data based on the coronary artery apocalypse data on the samples.
- the expansion module further includes: a noise expansion submodule, configured to add noise data to the blood vessel data in the region to be identified, where the noise data includes plaque data and artifact data; a false positive expansion submodule, configured to add noise data to the blood vessel data in the region to be identified; Add false positive data to the identified area vessel data.
- a noise expansion submodule configured to add noise data to the blood vessel data in the region to be identified, where the noise data includes plaque data and artifact data
- a false positive expansion submodule configured to add noise data to the blood vessel data in the region to be identified
- the identification module further includes: an output sub-module, configured to train a neural network model with the samples with the blood vessel expansion data and coronary apocalypse data to be identified to obtain coronary artery segmentation data; a loss sub-module, configured to The loss result is obtained based on the coronary artery segmentation data and the coronary artery identification data; the adjustment sub-module is configured to adjust the parameters of the neural network model based on the loss result.
- the sample acquisition module further includes: an original image acquisition sub-module, configured to acquire an original image, the original image containing the identification of the blood vessels and coronary arteries to be identified; and a preprocessing sub-module, configured to perform sliding window processing on the original image , to acquire multiple samples.
- an original image acquisition sub-module configured to acquire an original image, the original image containing the identification of the blood vessels and coronary arteries to be identified
- a preprocessing sub-module configured to perform sliding window processing on the original image , to acquire multiple samples.
- an embodiment of the present application provides an electronic device, including: a processor; a memory; and computer program instructions stored in the memory, where the computer program instructions are processed When executed, the processor causes the processor to perform a coronary artery segmentation method as described in any of the above.
- an embodiment of the present application provides a computer-readable storage medium, where computer program instructions are stored on the computer-readable storage medium, and the computer program instructions, when executed by a processor, cause all The processor executes the coronary artery segmentation method as described in any preceding one.
- the coronary artery segmentation method obtains the regional coronary artery segmentation data of the to-be-detected region by inputting the to-be-detected region with regional coronary artery enlightenment data into the above-mentioned coronary artery segmentation model; compares the regional coronary artery segmentation Whether the data and the coronary apocalypse data are divided into more segmentation areas, obtain the comparison result, replace the area to be detected in the image to be detected with the comparison result, obtain the replaced image to be detected, complete the detection step, and repeat the detection step, Select a new area to be detected in the replaced image to be detected, use the partial comparison result in the previous detection step as the coronary enlightenment data in the next detection step, and iterate repeatedly until all the images to be detected are segmented by regional coronary arteries After the data is traversed and replaced, the coronary artery segmentation results are obtained.
- Part of the segmentation structure in the previous detection step is used as the regional coronary enlightenment data in the next detection step, which provides inspiration for the segmentation in the next detection step, and makes full use of the connectivity of coronary vessel growth to achieve accurate segmentation and effective at the same time. Distinguish coronary and venous false positives and is robust to artifacts or plaques.
- the training method of the coronary artery segmentation model mentioned in the coronary artery segmentation method provided in the embodiment of the present application obtains the expansion data of the blood vessel to be identified by adding interference item data to the blood vessel data of the region to be identified in the sample;
- the identification data is processed to obtain the coronary apocalypse data that provides inspiration for the blood vessel data in the region to be identified during segmentation;
- the samples with the to-be-identified blood vessel expansion data, coronary artery apocalypse data and coronary artery identification data are input into the neural network model for training,
- the neural network model is enabled to output coronary identification data based on cues from coronary apocalypse data on the sample.
- the neural network model is trained by this sample, so that the neural network model learns how to distinguish the interference and achieve accurate segmentation.
- the coronary apocalypse data also provides partial enlightenment.
- the neural network model is trained through this sample, so that the neural network model learns how to obtain all the coronary artery segmentation results based on the partial enlightenment.
- FIG. 1 is a schematic flowchart of a method for training a coronary artery segmentation model according to an embodiment of the present application.
- FIG. 2 shows a schematic flowchart of acquiring expanded data of blood vessels to be identified in a method for training a coronary artery segmentation model according to an embodiment of the present application.
- FIG. 3 shows a schematic flowchart of obtaining coronary artery enlightenment data in a method for training a coronary artery segmentation model according to an embodiment of the present application.
- FIG. 4 shows a schematic flowchart of training a neural network model in a method for training a coronary artery segmentation model according to an embodiment of the present application.
- FIG. 5 is a schematic flowchart of sample acquisition in a method for training a coronary artery segmentation model according to an embodiment of the present application.
- FIG. 6 is a schematic flowchart of a coronary artery segmentation method according to an embodiment of the present application.
- FIG. 7 shows a schematic flowchart of acquiring the to-be-detected region and regional coronary artery enlightenment data when the to-be-detected region is the coronary origin region in a coronary artery segmentation method according to an embodiment of the present application.
- FIG. 8 shows the acquisition of the to-be-detected area and the regional coronary enlightenment when the to-be-detected area comes from the replaced to-be-detected image in the previous detection step of the current detection step in a coronary artery segmentation method provided by an embodiment of the present application Schematic diagram of the data flow.
- FIG. 9 is a schematic flowchart of obtaining a comparison result in a method for segmenting a coronary artery according to an embodiment of the present application.
- FIG. 10 is a schematic flowchart of a coronary artery segmentation method according to an embodiment of the present application.
- FIG. 11 is a schematic structural diagram of a training apparatus for a coronary artery segmentation model according to an embodiment of the present application.
- FIG. 12 is a schematic structural diagram of a training apparatus for a coronary artery segmentation model according to an embodiment of the present application.
- FIG. 13 is a schematic structural diagram of a coronary artery segmentation device according to an embodiment of the present application.
- FIG. 14 is a schematic structural diagram of a coronary artery segmentation device according to an embodiment of the present application.
- FIG. 15 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
- Deep learning realizes artificial intelligence in computing systems by building artificial neural networks with hierarchical structures. Since the hierarchically structured artificial neural network can extract and filter the input information layer by layer, deep learning has the capability of representation learning and can realize end-to-end supervised learning and unsupervised learning.
- the hierarchical artificial neural network used in deep learning has various forms, and the complexity of its hierarchy is commonly referred to as "depth". According to the type of construction, the form of deep learning includes multilayer perceptrons, convolutional neural networks, and recurrent neural networks. , Deep Belief Networks, and other hybrid constructs. Deep learning uses data to update the parameters in its construction to achieve training goals. This step is generally called "learning”. Deep learning proposes a method for computers to automatically learn pattern features, and incorporate feature learning into building models. steps, thereby reducing the incompleteness caused by artificial design features.
- Neural network is an operation model, which is composed of a large number of nodes (or neurons) connected to each other, each node corresponds to a strategy function, and the connection between each two nodes represents a weighted value for the signal passing through the connection, Call it weight.
- the neural network generally includes multiple neural network layers, the upper and lower network layers are cascaded with each other, the output of the i-th neural network layer is connected to the input of the i+1-th neural network layer, and the output of the i+1-th neural network layer is connected. Connect to the input of the i+2th neural network layer, and so on.
- each neural network layer After the training samples are input to the cascaded neural network layers, each neural network layer outputs an output result, and the output result is used as the input of the next neural network layer. Thus, the output is obtained through multiple neural network layer calculations, and the output layer is compared. The output prediction result and the real target value, and then adjust the weight matrix and strategy function of each layer according to the difference between the prediction result and the target value.
- the neural network uses the training samples to continuously go through the above adjustment steps, so that the neural network The weights and other parameters of the neural network are adjusted until the predicted results output by the neural network are consistent with the real target results. This step is called the training step of the neural network. After the neural network is trained, a neural network model can be obtained.
- Cardiovascular disease is one of the diseases with the highest morbidity and mortality in the world today.
- Coronary CTA Computed Tomography Angiography, CT Angiography
- CT Computerputed Tomography, Computed Tomography
- the segmentation of cardiac coronary arteries is a difficult problem.
- the specific reasons are as follows: the structure of the coronary arteries is complex, and there are many small blood vessels with many branches; the grayscale of the coronary arteries is uneven, the contrast with the surrounding tissues is low, and the boundary of the peripheral part of the blood vessels is blurred; the coronary arteries contain various complex lesions; Motion artifacts will affect the imaging of coronary arteries; there are many veins in the heart interlaced with coronary arteries, and when the image quality is not high, there is a phenomenon that veins are connected to coronary arteries, and segmented vein false positives are prone to occur.
- the coronary artery segmentation method obtains the regional coronary artery segmentation data of the to-be-detected region by inputting the to-be-detected region with regional coronary artery enlightenment data into the above-mentioned coronary artery segmentation model; compares the regional coronary artery segmentation Whether the data and coronary apocalypse data are divided into more segmentation areas, obtain the comparison result, replace the area to be detected in the image to be detected with the comparison result, obtain the replaced image to be detected, in the replaced image to be detected Select a new area to be detected, take part of the comparison results in the previous detection step as the coronary enlightenment data in the next detection step, repeat the detection step, and iterate repeatedly until all the images to be detected are traversed and replaced by the regional coronary artery segmentation data.
- the embodiment of the present application also provides a training method for the coronary artery segmentation model.
- Data; coronary artery apocalypse data that provides inspiration for the blood vessel data in the region to be identified during segmentation is obtained by processing the coronary artery identification data of the sample; the sample with the expansion data of the blood vessel to be identified, the coronary artery apocalypse data and the coronary artery identification data
- the neural network model is input for training, so that the neural network model can output coronary artery identification data based on the prompt of the coronary apocalypse data on the sample.
- the neural network model is trained by this sample, so that the neural network model learns how to distinguish the interference and achieve accurate segmentation.
- the coronary apocalypse data also provides partial enlightenment.
- the neural network model is trained through this sample, so that the neural network model learns how to obtain all the coronary artery segmentation results based on the partial enlightenment.
- FIG. 1 is a schematic flowchart of a method for training a coronary artery segmentation model according to an embodiment of the present application. As shown in Figure 1, the training method of the coronary artery segmentation model includes the following steps.
- Step 101 Obtain a sample, where the sample includes blood vessel data and coronary artery identification data in the region to be identified.
- the sample is a part of the identified cardiac CT image
- the identified cardiac CT image includes the blood vessel to be identified and the coronary artery identification that has been segmented from the coronary artery. Since the sample is a part of the identified cardiac CT image, the sample includes the vessel data of the region to be identified and the coronary artery identification data.
- the coronary artery identification data is the identified and segmented coronary vessels.
- the embodiment of the present application does not specifically limit the segmentation method.
- the specific marking method is not limited in the embodiment of the present application, whether it is manually marked manually or marked by other methods.
- Step 102 adding interference data to the blood vessel data in the region to be identified, and acquiring the expanded data of the blood vessel to be identified.
- the interference data is added to the blood vessel data in the to-be-identified region to obtain the expanded data of the blood vessels to be identified. Then, the neural network model is trained through the expanded data of the blood vessels to be identified, so that the neural network model has the ability to reject interference and accurately segment.
- Step 103 Obtain coronary artery enlightenment data based on the coronary artery identification data, and the coronary artery enlightenment data is used to provide enlightenment when identifying the blood vessel data in the region to be identified.
- the coronary artery segmentation model in the prior art uses samples to train the neural network model, and uses the identified coronary blood vessels in the samples as the output target, so that the neural network model learns how to identify the coronary arteries based on the blood vessels to be identified.
- the segmentation results of coronary vessels output by the trained neural network model may be different from the identified coronary vessels, and the robustness of the output of the neural network model is low.
- the coronary apocalypse data is a part of the identified blood vessels to be identified.
- it not only provides the neural network model with the blood vessel data in the region to be identified and the coronary artery identification data as the output target.
- the neural network model such as the source of some of the blood vessels that have been segmented, making full use of the connectivity of blood vessel growth to make the segmentation results output by the neural network model more accurate.
- the coronary apocalypse data is obtained based on the coronary artery identification data to provide an accurate origin point for the neural network model as the inspiration.
- Step 104 Input the sample with the blood vessel expansion data to be identified, the coronary artery apocalypse data and the coronary artery identification data into the neural network model, and train the neural network model so that the neural network model can output coronary artery based on the coronary artery apocalypse data on the sample.
- Arterial identification data Input the sample with the blood vessel expansion data to be identified, the coronary artery apocalypse data and the coronary artery identification data into the neural network model, and train the neural network model so that the neural network model can output coronary artery based on the coronary artery apocalypse data on the sample. Arterial identification data.
- acquiring the coronary apocalypse data based on the coronary artery identification data includes: selecting a portion from the coronary artery identification data as the coronary artery apocalypse data.
- the neural network model is input with the samples with the expansion data of the blood vessels to be identified, the coronary artery enlightenment data and the coronary artery identification data, the enlightenment factors provided by the coronary artery enlightenment data are used, and the coronary artery identification data is used as the output target to make the neural network model
- Obtaining rejection interference items can also be accurately segmented under the premise of artifacts or plaques, enabling the neural network model to obtain vessel-based connectivity and output all accurate coronary artery segmentation results according to partial prompts.
- the blood vessel expansion data to be identified is obtained; by processing the coronary artery identification data of the sample, the obtained blood vessel data provides the blood vessel data of the to-be-identified area during segmentation.
- the inspired coronary apocalypse data; the samples with the vessel expansion data to be identified, the coronary artery apocalypse data and the coronary artery identification data are input into the neural network model for training, so that the neural network model can output the coronary artery based on the prompt of the coronary apocalypse data on the sample.
- Arterial identification data is input into the neural network model for training, so that the neural network model can output the coronary artery based on the prompt of the coronary apocalypse data on the sample.
- the neural network model is trained by this sample, so that the neural network model learns how to distinguish the interference and achieve accurate segmentation.
- the coronary apocalypse data also provides partial enlightenment.
- the neural network model is trained through this sample, so that the neural network model learns how to obtain all the coronary artery segmentation results based on the partial enlightenment.
- FIG. 2 shows a schematic flowchart of acquiring expanded data of blood vessels to be identified in a method for training a coronary artery segmentation model according to an embodiment of the present application.
- adding interference data to the blood vessel data in the to-be-identified region, and acquiring the expanded data of the to-be-identified blood vessel includes the following steps.
- Step 2021 Add noise data to the blood vessel data in the region to be identified, where the noise data includes plaque data and artifact data.
- the noise such as plaque data and artifact data should affect the robustness of coronary artery segmentation.
- Noise data is added to the blood vessel data in the region to be identified to train the neural network model.
- the neural network model has the ability to distinguish interfering factors to achieve accurate segmentation.
- Step 2022 Add false positive data to the blood vessel data in the region to be identified.
- the neural network model is difficult to distinguish, and the coronary veins are mistaken as the target, forming a vein false positive. False positive data is added to the blood vessel data in the region to be identified to train the neural network model, so that the neural network model has the ability to distinguish coronary veins from coronary arteries.
- a specific method for adding false positive data to the blood vessel data in the to-be-identified region may be to fuse the blood vessel data of other regions into the blood vessel data of the to-be-identified region, but avoid the identified coronary vessels represented by the coronary artery identification data. It should be understood that as long as the blood vessel data of other regions can be fused to the blood vessel data of the to-be-identified region, the present application does not limit the specific fusion means.
- Fig. 3 shows a schematic flowchart of obtaining coronary revelation data in a method for training a coronary artery segmentation model according to an embodiment of the present application.
- obtaining coronary apocalypse data based on coronary artery identification data includes the following steps.
- Step 3031 Select a part from the coronary artery identification data as coronary apocalypse data.
- the selected part of the coronary artery identification data is used as the coronary artery enlightenment data to provide a basis for subsequent blood vessel segmentation.
- the selected part of the coronary artery identification data is used as the coronary artery enlightenment data to provide a basis for subsequent blood vessel segmentation.
- FIG. 4 shows a schematic flowchart of training a neural network model in a method for training a coronary artery segmentation model according to an embodiment of the present application.
- the samples with the expansion data of the blood vessels to be identified and the coronary apocalypse data are input into the neural network model. Carrying out the training involves the following steps.
- Step 4041 Train the neural network model with the samples with the blood vessel expansion data to be identified and the coronary artery apocalypse data to obtain coronary artery segmentation data.
- a sample is input in the neural network model, and coronary artery segmentation data is output, and the coronary artery segmentation data is the segmentation prediction result of the expanded data of blood vessels to be identified by the neural network model.
- Step 4042 Obtain a loss result based on the coronary artery segmentation data and the coronary artery identification data.
- the coronary artery identification data is an input reference value
- the coronary artery segmentation data is an output value. There is a difference between the output value and the input reference value, and a loss result between the two is obtained.
- Step 4043 Based on the loss result, adjust the parameters of the neural network model.
- the loss result in the preset range indicates that the difference between the coronary artery segmentation data and the coronary artery identification data is within the preset range.
- the sample is input into the neural network model to obtain coronary artery segmentation data
- the coronary artery segmentation data is the input reference value
- the coronary artery segmentation data is the output value
- the loss result between the two is obtained
- the parameters of the neural network model are adjusted. , until the loss result is within the preset range, stop adjusting the parameters of the neural network model. Improve the ability of the model to accurately segment.
- FIG. 5 is a schematic flowchart of sample acquisition in a method for training a coronary artery segmentation model according to an embodiment of the present application. As shown in Figure 5, acquiring a sample includes the following steps.
- Step 5001 Obtain an original image, where the original image contains the identification of the blood vessel and coronary artery to be identified.
- the original image is a cardiac CT image
- the cardiac CT image includes the blood vessel to be identified and the identified coronary artery standard
- Step 5002 Perform sliding window processing on the original image to obtain multiple samples.
- multiple samples are obtained by performing sliding window processing on the original image, and the original image is divided into multiple samples for processing respectively, so as to improve the deep learning efficiency of the neural network model.
- FIG. 6 is a schematic flowchart of a coronary artery segmentation method according to an embodiment of the present application. As shown in Fig. 6, the coronary artery segmentation method includes the following steps.
- Step 601 Obtain the region to be detected and regional coronary apocalypse data, and the regional coronary apocalypse data is used to provide enlightenment when the region to be detected is segmented.
- the image to be detected can be divided into a plurality of regions to be detected according to different selection rules.
- the entire segmentation process can include multiple detection steps.
- the detection step is the first step in the entire segmentation process
- the region to be detected is the coronary origin region.
- the detection step is not the first step in the entire segmentation process
- the region to be detected is from the origin of the coronary artery.
- the regional coronary enlightenment data is to provide enlightenment data for outputting the segmentation result corresponding to the to-be-detected region when the coronary artery segmentation model segments the to-be-detected image.
- Step 602 Input the region to be detected with regional coronary arterial enlightenment data into the coronary artery segmentation model trained by any of the above methods, and obtain regional coronary artery segmentation data.
- the coronary artery segmentation model trained by any of the above methods has already obtained the coronary artery segmentation results of all regions based on partial region enlightenment, inputting the region to be detected with regional coronary artery enlightenment data into the coronary artery segmentation model, it is possible to Obtain regional coronary artery segmentation data corresponding to the region to be detected.
- the regional coronary artery segmentation data is the output result of the region corresponding to the region to be detected.
- Step 603 Obtain a comparison result between the regional coronary apocalypse data and the regional coronary artery segmentation data.
- the regional coronary apocalypse data is the result of partial enlightenment and is not complete and accurate.
- the regional coronary artery segmentation data output by the coronary artery segmentation model is a more accurate segmentation result that excludes interference and false positives.
- Regional coronary artery segmentation data is more accurate, but considering factors such as image area selection, it is necessary to obtain the comparison results of regional coronary apocalypse data and regional coronary artery segmentation data to extract more accurate segmentation results.
- Step 604 Replace the to-be-detected area in the to-be-detected image with the comparison result, and acquire the replaced to-be-detected image to complete the detection step.
- Step 605 Repeat the detection step, wherein, for a plurality of detection steps corresponding to the image to be detected, select the area to be detected in the next detection step from the replaced image to be detected in the previous detection step, and select a part of the detection step from the previous detection step.
- the comparison result in is used as the coronary enlightenment data in the next detection step.
- the entire segmentation process may include a plurality of detection steps.
- the above steps 601 to 604 are one detection step.
- the area to be detected in the next detection step is selected from the replaced image to be detected in the previous detection step, and part of the previous detection step is selected.
- the comparison result in the step is used as the coronary revelation data in the next detection step, making full use of the connectivity of blood vessels to achieve accurate segmentation and improve the robustness of segmentation.
- the replaced images to be detected obtained in the last detection step are output as coronary artery segmentation data.
- the coronary artery segmentation method obtains the regional coronary artery segmentation data of the to-be-detected region by inputting the to-be-detected region with regional coronary artery enlightenment data into the above-mentioned coronary artery segmentation model; compares the regional coronary artery segmentation Whether the data and coronary apocalypse data are divided into more segmentation areas, obtain the comparison result, replace the area to be detected in the image to be detected with the comparison result, obtain the replaced image to be detected, in the replaced image to be detected Select a new area to be detected, take part of the comparison results in the previous detection step as the coronary enlightenment data in the next detection step, repeat the detection step, and iterate repeatedly until all the images to be detected are traversed and replaced by the regional coronary artery segmentation data.
- FIG. 7 shows a schematic flowchart of acquiring the to-be-detected region and regional coronary artery enlightenment data when the to-be-detected region is the coronary origin region in a coronary artery segmentation method according to an embodiment of the present application.
- acquiring the region to be detected and the regional coronary apocalypse data includes the following steps.
- Step 7011 Based on the image to be detected, obtain the segmentation data of the origin of the coronary artery.
- the image to be detected is input into the trained origin segmentation module, and the segmentation result of the connection area between the aorta and the coronary artery is obtained.
- the coronary origin segmentation data can provide key features for subsequent segmentation.
- Step 7012 Extract the first skeleton from the coronary artery origin segmentation data.
- a first skeleton representing the contour and orientation of the coronary arteries is extracted.
- Step 7013 Select the first seed point on the first skeleton according to the first preset step size.
- the first preset step size is limited according to a specific application scenario, and the value of the first preset step size is not limited in this embodiment of the present application.
- Step 7014 Push the first seed into the seed point stack, and obtain the seed point at the top of the first stack.
- Step 7015 Taking the center of the seed point at the top of the first stack as the center, on the image to be detected, select a first area of a first preset size as the area to be detected.
- the size of the first preset size is selected according to the size of the image to be detected. For example, when the size of the large detected image is 512 ⁇ 512 ⁇ 512 pixels, the size of the first preset size can be 128 ⁇ 128 ⁇ 128 pixels .
- the first seed point is selected on the first skeleton according to the first preset step size, the first seed is pressed into the seed point stack, and the seed point at the top of the first stack is obtained, and the center of the seed point at the top of the first stack is taken as the center In the center, on the image to be detected, a first area with a first preset size is selected as the area to be detected, and the area to be detected is acquired through the above operations.
- Step 7016 Use the coronary artery origin segmentation data in the first region as regional coronary artery enlightenment data.
- the coronary origin segmentation data in the first region can provide inspiration for segmenting the region to be detected, so that the coronary artery segmentation model fully considers the vascular connectivity, and outputs Accurate regional coronary artery segmentation data corresponding to the region to be detected.
- the segmentation result of the connection between the coronary artery and the aorta is obtained through the origin segmentation module, and the first seed point is generated and pushed into the seed point stack. Then pop the top seed point of the first stack, and predict a new segmentation based on the obtained segmentation data of the first region and the origin of the coronary artery in the first region, so that the coronary artery segmentation model fully considers the blood vessel connectivity, and the output is the same as the one to be Accurate regional coronary artery segmentation data corresponding to the detection region.
- FIG. 8 shows the acquisition of the to-be-detected area and the regional coronary enlightenment when the to-be-detected area comes from the replaced to-be-detected image in the previous detection step of the current detection step in a coronary artery segmentation method provided by an embodiment of the present application Schematic diagram of the data flow.
- the to-be-detected area comes from the replaced to-be-detected image in the previous detection step of the current detection step.
- acquiring the to-be-detected area and the regional coronary apocalypse data includes the following steps.
- Step 8061 Obtain newly added coronary artery segmentation data from the comparison result in the previous detection step.
- the newly added coronary artery segmentation data is obtained from the comparison result of the regional coronary artery segmentation data and the regional coronary artery enlightenment data in the previous detection step.
- the newly added coronary artery segmentation data is a newly segmented region obtained by comparing the regional coronary artery segmentation data with the regional coronary artery enlightenment data.
- Step 8062 Extract the second skeleton in the newly added coronary artery segmentation data.
- the second skeleton in the extraction of the newly added coronary artery segmentation data is added, and a part of the area to be detected that is obtained subsequently belongs to the segmentation that has already been obtained. As a result, there is also another part belonging to the new region to be segmented extending in the direction of the blood flow.
- the second skeleton is extracted based on the contour and orientation of the coronary artery in the newly added coronary artery data.
- Step 8063 Select the second seed point on the second skeleton according to the second preset step size.
- Step 8064 Push the second seed point into the seed point stack, and obtain the seed point at the top of the second stack.
- Step 8065 Taking the center of the seed point at the top of the second stack as the center, on the replaced image to be detected in the previous detection step, select the second area of the second preset size as the area to be detected, and the second area covers Part of the alignment results from the previous detection step.
- step 8063, step 8064 and step 8065 the method for obtaining the second area based on the second seed point is similar to the method for obtaining the first area based on the first seed point, and details are not described here.
- the second region needs to cover a part of the comparison result in the previous detection step, so that a part of the segmentation result of the previous detection step can be sent into the coronary artery segmentation model to provide inspiration for the detection.
- Step 8066 Use the comparison result in the previous detection step in the second region as the regional coronary apocalypse data.
- the comparison result in the previous detection step in the second area is used as the accurate output of the previous detection step, which is the current area to be detected. Segmentation provides inspiration.
- the coronary artery segmentation data is output.
- FIG. 9 is a schematic flowchart of obtaining a comparison result in a method for segmenting a coronary artery according to an embodiment of the present application. As shown in FIG. 9 , obtaining the comparison result between the regional coronary artery apocalypse data and the regional coronary artery segmentation data includes the following steps.
- Step 9031 When the regional coronary artery segmentation data has more segmented regions than the regional coronary apocalypse data, select the regional coronary artery segmentation data as the comparison result. or,
- Step 9032 When the regional coronary artery segmentation data does not segment more segmented regions than the regional coronary apocalypse data, select the regional coronary artery enlightenment data as the comparison result.
- the regional coronary artery segmentation data segmented by the coronary artery segmentation model trained by the above training method has higher accuracy, but due to regional vascular factors and model factors, there may be regional coronary artery segmentation data that is the same as the coronary apocalypse data.
- the regional coronary artery segmentation data segmented more segmentation regions than the regional coronary artery enlightenment data
- the regional coronary artery segmentation data was selected as the comparison result
- the regional coronary artery segmentation data did not segment more than the regional coronary artery enlightenment data.
- There are many segmentation regions and the regional coronary apocalypse data is selected as the comparison result.
- FIG. 10 is a schematic flowchart of a coronary artery segmentation method according to an embodiment of the present application.
- a cardiac CT image as the image to be detected, input the image to be detected into the trained origin segmentation module, segment the connection area between the aorta and the coronary artery, and obtain the segmentation data of the origin of the coronary artery (refer to step 10011 in FIG. 10 ). ).
- a first skeleton representing the contour and orientation of the coronary artery is extracted from the coronary artery origin segmentation data (refer to step 10012 in FIG. 10 ).
- a first area with a first preset size is selected as the first area to be detected (refer to steps 10013 , 10014 and 10015 in FIG. 10 ).
- the coronary artery origin segmentation data in the first region is used as the first region coronary arterial enlightenment data (refer to step 10016 in FIG. 10 ).
- the first region coronary artery segmentation data corresponding to the region to be detected can be obtained (refer to step 1002 in FIG. 10 ).
- the coronary artery segmentation data in the first region divides more segmentation regions than the coronary artery enlightenment data in the first region
- the coronary artery segmentation data in the first region is selected as the comparison result (refer to step 9031 in FIG.
- the coronary artery segmentation data in one region does not segment more segmented regions than the coronary artery enlightenment data in the first region, and the coronary artery enlightenment data in the first region is selected as the comparison result (refer to step 9032 in FIG. 10 ).
- the first to-be-detected area in the to-be-detected image is replaced with the comparison result, the replaced to-be-detected image is acquired (refer to step 1004 in FIG. 10 ), and the first detection step is completed.
- the newly added coronary artery segmentation data is obtained from the comparison result in the previous detection step (refer to step 10061 in FIG. 10 ).
- the second skeleton is extracted from the newly added coronary artery segmentation data (refer to step 10062 in FIG. 10 ).
- a second area of the second preset size is selected as the second to-be-detected area, and the second area covers part of the comparison result in the previous detection step (refer to Fig.
- Step 10063, Step 10064 and Step 10065 in 10 Since the second region needs to cover a part of the comparison result in the previous detection step, the comparison result in the previous detection step in the second region is used as the coronary apocalypse data in the second region (refer to step 10066 in FIG. 10 ). ); the second region to be detected with the second region coronary apocalypse data is input into the coronary artery segmentation model to obtain the second region coronary artery segmentation data (refer to step 1002 in FIG.
- the coronary artery segmentation data is output (refer to step 1005 in FIG. 10 ).
- FIG. 11 is a schematic structural diagram of a training apparatus for a coronary artery segmentation model according to an embodiment of the present application.
- the training device 1100 of the coronary artery segmentation model includes:
- the sample acquisition module 1101 is configured to acquire samples, including the blood vessel data of the region to be identified and the coronary artery identification data; the expansion module 1102 is configured to add interference data to the blood vessel data of the to-be-identified region to obtain the expanded data of the blood vessels to be identified;
- Enlightenment acquisition module 1103, is configured to obtain coronary apocalypse data based on coronary artery identification data, and the coronary artery apocalypse data is used to provide enlightenment when identifying the blood vessel data in the region to be identified, wherein acquiring coronary artery apocalypse data based on the coronary artery identification data includes: from coronary artery identification data.
- the selected part of the arterial identification data is used as coronary apocalypse data; and the identification module 1104 is configured to input the samples with the vessel expansion data to be identified, the coronary arterial apocalypse data and the coronary artery identification data into the neural network model, and the coronary artery identification data is the output
- the objective is to train the neural network model so that the neural network model can output coronary artery identification data based on the coronary apocalypse data on the sample.
- the expansion module 1102 is used to add interference item data to the blood vessel data of the region to be identified in the sample to obtain the expanded data of the blood vessel to be identified;
- the revelation acquisition module 1103 is used to process the coronary artery identification data of the sample to obtain when segmentation is performed Coronary artery enlightenment data that provides enlightenment for the blood vessel data of the region to be identified;
- the identification module 1104 inputs the samples with the vessel expansion data to be identified, coronary artery enlightenment data and coronary artery identification data into the neural network model for training, so that the neural network model can be based on the sample.
- Coronary artery identification data is output on the prompt of the coronary apocalypse data.
- the neural network model is trained by this sample, so that the neural network model learns how to distinguish the interference and achieve accurate segmentation.
- the coronary apocalypse data also provides partial enlightenment.
- the neural network model is trained through this sample, so that the neural network model learns how to obtain all the coronary artery segmentation results based on the partial enlightenment.
- FIG. 12 is a schematic structural diagram of a training apparatus for a coronary artery segmentation model according to an embodiment of the present application.
- the expansion module 1102 includes: a noise expansion sub-module 11021, which is configured to add noise data to the blood vessel data in the region to be identified, and the noise data includes plaque data and artifact data; and a false positive expansion sub-module 11022, which is False positive data is added to the blood vessel data in the region to be identified.
- the enlightenment obtaining module 1103 is further configured to select a portion from the coronary artery identification data as the coronary enlightenment data.
- the identification module 1104 includes: an output sub-module 11041, configured to train a neural network model with samples with the to-be-identified vessel expansion data and coronary apocalypse data to obtain coronary artery segmentation data
- the loss sub-module 11042 is configured to obtain the loss result based on the coronary artery segmentation data and the coronary artery identification data
- the adjustment sub-module 11043 is configured to adjust the parameters of the neural network model based on the loss result.
- the sample acquisition module 1101 further includes: an original image acquisition sub-module 1105 , configured to acquire an original image, the original image including the identification of the blood vessel and coronary artery to be identified; and a preprocessing sub-module 1106 , configured to perform sliding window processing on the original image to acquire multiple samples.
- an original image acquisition sub-module 1105 configured to acquire an original image, the original image including the identification of the blood vessel and coronary artery to be identified
- a preprocessing sub-module 1106 configured to perform sliding window processing on the original image to acquire multiple samples.
- FIG. 13 is a schematic structural diagram of a coronary artery segmentation device according to an embodiment of the present application.
- the coronary artery segmentation device 1300 includes: a segmentation acquisition module 1301 configured to acquire a region to be detected and regional coronary apocalypse data, where the regional coronary apocalypse data is used to provide enlightenment when segmenting the region to be detected Segmentation module 1302, configured to input the region to be detected with regional coronary artery apocalypse data into the coronary artery segmentation model trained using the training method of any of the above-mentioned coronary artery segmentation models to obtain regional coronary artery segmentation data;
- the module 1303 is configured to obtain the comparison result of the regional coronary artery segmentation data and the regional coronary apocalypse data;
- the replacement module 1304 is configured to replace the to-be-detected area in the to-be-detected image with the comparison result, and obtain the replaced to
- the coronary artery segmentation data is output.
- the coronary artery segmentation module 1302 obtains the regional coronary artery segmentation data of the to-be-detected region by inputting the region to be detected with the regional coronary apocalypse data into the above-mentioned coronary artery segmentation model; Compare whether the regional coronary artery segmentation data and the coronary apocalypse data are divided into more segmentation regions, and obtain the comparison result.
- the replacement module 1304 replaces the to-be-detected area in the to-be-detected image with the comparison result, and obtains the replaced to-be-detected image
- the output module 1305 repeats the detection step, and iterates repeatedly until all the images to be detected are traversed and replaced by the regional coronary artery segmentation data, and obtain the coronary artery segmentation result.
- the coronary artery segmentation device uses the partial segmentation structure in the previous detection step as the regional coronary enlightenment data in the next detection step, provides inspiration for segmentation in the next detection step, and makes full use of the growth of coronary vessels It can achieve accurate segmentation while effectively distinguishing coronary arteries and vein false positives, and has strong robustness to artifacts or plaques.
- FIG. 14 is a schematic structural diagram of a coronary artery segmentation device according to an embodiment of the present application.
- the region to be detected is the region of origin of the coronary artery
- the segmentation and acquisition module 1301 includes: an origin segmentation sub-module 13011, which is configured to acquire segmentation data of the origin of the coronary artery based on the image to be detected; the first extraction sub-module 13022, which is configured as Extract the first skeleton from the coronary artery origin segmentation data; the first seed point selection sub-module 13023 is configured to select the first seed point on the first skeleton according to the first preset step size; the first stack top seed point acquisition sub-module 13024, configured to push the first seed into the seed point stack, and obtain the first stack top seed point; the first area acquisition sub-module 13025, configured to take the center of the first stack top seed point as the center, on the image to be detected A first region with a first preset size is selected as the region to be detected; and the
- the area to be detected is from the replaced image to be detected in the previous detection step of the current detection step
- the segmentation and acquisition module 1301 further includes a newly added acquisition sub-module 13061, which is configured to The newly added coronary artery segmentation data is obtained from the comparison result in the previous detection step
- the second extraction sub-module 13062 is matched with the second skeleton in the extraction of the newly-added coronary artery segmentation data
- the second seed point selection sub-module 13063 is configured In order to select the second seed point according to the second preset step size on the second skeleton
- the second stack top seed point acquisition submodule 13064 is configured to push the second seed point into the seed point stack, and obtain the second stack top seed
- the second area acquisition submodule 13065 is configured to take the center of the second stack top seed point as the center, and select the second area of the second preset size on the replaced image to be detected as the second area in the next detection step.
- the second area covers part of the comparison result on the replaced image to be detected in the previous detection step, selects the second area of the second preset size as the area to be detected, and the second area covers part of the previous image.
- the comparison result in the detection step; the second enlightenment obtaining sub-module 13066 is configured to use the comparison result in the previous detection step in the second region as the regional coronary artery enlightenment data.
- the comparison and determination module 1302 is further configured to select the regional coronary artery segmentation data as the comparison result when the regional coronary artery segmentation data divides more segmented regions than the regional coronary arterial enlightenment data; or when the regional coronary artery segmentation data is the comparison result; The segmented data did not segment more segmented regions than the regional coronary apocalypse data, and the regional coronary apocalypse data was selected as the comparison result.
- FIG. 15 is a schematic structural diagram of an electronic device according to an embodiment of the present application. As shown in FIG. 15 , electronic device 1500 includes one or more processors 1510 and memory 1520 .
- Processor 1510 may be a central processing unit (CPU) or other form of processing unit having data processing capabilities and/or instruction execution capabilities, and may control other components in electronic device 1500 to perform desired functions.
- CPU central processing unit
- Processor 1510 may be a central processing unit (CPU) or other form of processing unit having data processing capabilities and/or instruction execution capabilities, and may control other components in electronic device 1500 to perform desired functions.
- Memory 1520 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and/or non-volatile memory.
- the volatile memory may include, for example, random access memory (RAM) and/or cache memory, or the like.
- the non-volatile memory may include, for example, read only memory (ROM), hard disk, flash memory, and the like.
- One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 1510 may execute the program instructions to implement the method for training the coronary artery segmentation model of the various embodiments of the present application described above and coronary artery segmentation methods and/or other desired features.
- the electronic device 1500 may also include an input device 1530 and an output device 1540 interconnected by a bus system and/or other form of connection mechanism (not shown).
- the input device 1530 may be the above-mentioned microphone or microphone array for capturing the input signal of the sound source.
- the input device 1530 may be a communication network connector.
- the input device 1530 may also include, for example, a medical image acquisition device or the like.
- the output device 1540 can output various information to the outside, including the determined target object information and the like.
- the output devices 1540 may include, for example, displays, speakers, printers, and communication networks and their connected remote output devices, among others.
- the electronic device 1500 may also include any other suitable components according to the specific application.
- embodiments of the present application may also be computer program products comprising computer program instructions that, when executed by a processor, cause the processor to perform the "Example Routine Coronary Artery” described above in this specification.
- the computer program product can write program codes for performing the operations of the embodiments of the present application in any combination of one or more programming languages, including object-oriented programming languages, such as Java, C++, etc. , also includes conventional step-by-step programming languages, such as "C" language or similar programming languages.
- the program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device, or entirely on the remote computing device or server execute on.
- embodiments of the present application may also be computer-readable storage media on which computer program instructions are stored, the computer program instructions, when executed by a processor, cause the processor to perform the above-mentioned “Example Coronary Artery Segmentation” in this specification.
- the steps in the training method of the coronary artery segmentation model and the coronary artery segmentation method according to various embodiments of the present application are described in the "Model Training Method” and “Exemplary Coronary Artery Segmentation Method” sections.
- the computer-readable storage medium may employ any combination of one or more readable media.
- the readable medium may be a readable signal medium or a readable storage medium.
- the readable storage medium may include, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses or devices, or a combination of any of the above. More specific examples (non-exhaustive list) of readable storage media include: electrical connections with one or more wires, portable disks, hard disks, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disk read only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
- Apparatus For Radiation Diagnosis (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
Description
Claims (18)
- 一种冠状动脉分割方法,包括:获取待检测区域以及区域冠状动脉启示数据,所述区域冠状动脉启示数据用于在对所述待检测区域进行分割时提供启示;将带有所述区域冠状动脉启示数据的所述待检测区域输入到预先训练的冠状动脉分割模型,获取区域冠状动脉分割数据;获取所述区域冠状动脉分割数据与所述区域冠状动脉启示数据的比对结果;用所述比对结果替换待检测图像中的所述待检测区域,获取替换后的待检测图像,以完成检测步骤;重复所述检测步骤,其中,对于所述待检测图像对应的多个所述检测步骤,在上一个检测步骤中的所述替换后的待检测图像中选取下一个检测步骤的所述待检测区域,将上一个检测步骤中的所述比对结果的一部分作为下一个检测步骤中的所述冠状动脉启示数据;以及当所述待检测图像全部被所述比对结果遍历替换完,输出冠状动脉分割数据。
- 根据权利要求1所述的冠状动脉分割方法,其中,所述待检测区域为冠状动脉起源区域;其中,所述获取所述待检测区域以及所述区域冠状动脉启示数据包括:基于待检测图像,获取冠状动脉起源分割数据;在所述冠状动脉起源分割数据中提取第一骨架;在所述第一骨架上按照第一预设步长选取第一种子点;将所述第一种子压入种子点栈中,获取第一栈顶种子点;以所述第一栈顶种子点的中心为中心,在所述待检测图像上,选取第一预设尺寸大小的第一区域作为所述待检测区域;以及将所述第一区域内的所述冠状动脉起源分割数据作为所述区域冠状动脉启示数据。
- 根据权利要求1或2所述的冠状动脉分割方法,其中,所述待检测区域来自于当前检测步骤的上一个检测步骤中的替换后的待检测图像;其中,所述获取所述待检测区域以及所述区域冠状动脉启示数据包括:在所述上一个检测步骤中的比对结果中获取新增冠状动脉分割数据;在所述新增冠状动脉分割数据提取中第二骨架;在所述第二骨架上按照第二预设步长选取所述第二种子点;将所述第二种子点压入所述种子点栈中,获取第二栈顶种子点;以所述第二栈顶种子点的中心为中心,在所述上一个检测步骤中的替换后的待检测图像上,选取第二预设尺寸大小的第二区域作为所述待检测区域,所述第二区域覆盖部分所述上一个检测步骤中的比对结果;以及将所述第二区域内的所述上一个检测步骤中比对结果作为所述区域冠状动脉启示数据。
- 根据权利要求1至3中任一所述的冠状动脉分割方法,其中,所述获取所述区域冠状动脉分割数据与所述区域冠状动脉启示数据的比对结果包括:当所述区域冠状动脉分割数据比所述区域冠状动脉启示数据分割出更多的分割区域,选取所述区域冠状动脉分割数据为所述比对结果;或当所述区域冠状动脉分割数据比所述区域冠状动脉启示数据并未分割出更多的分割区域,选取所述区域冠状动脉启示数据为所述比对结果。
- 根据权利要求1至4中任一所述的所述的冠状动脉分割方法,其中,所述冠状动脉分割模型的训练方法包括:获取样本,所述样本包括待识别区域血管数据与冠状动脉标识数据;在所述待识别区域血管数据中增加干扰数据,获取待识别血管扩充数据;基于所述冠状动脉标识数据获取冠状动脉启示数据,所述冠状动脉启示数据用于在对所述 待识别区域血管数据进行标识时为神经网络模型提供一个准确的起源点提供启示;以及将带有所述待识别血管扩充数据、所述冠状动脉启示数据和所述冠状动脉标识数据的所述样本输入神经网络模型,对神经网络模型进行训练,以使得所述神经网络模型能够基于所述样本上的所述冠状动脉启示数据输出所述冠状动脉标识数据;其中,所述基于所述冠状动脉标识数据获取冠状动脉启示数据,包括:从所述冠状动脉标识数据中选取部分作为所述冠状动脉启示数据。
- 根据权利要求5所述的冠状动脉分割方法,其中,所述在所述待识别区域血管数据中增加干扰数据,获取待识别血管扩充数据包括:在所述待识别区域血管数据中添加噪声数据,所述噪声数据包括斑块数据和伪影数据;以及在所述待识别区域血管数据中添加假阳数据。
- 根据权利要求5或6所述的冠状动脉分割方法,其中,所述将带有所述待识别血管扩充数据、所述冠状动脉启示数据和所述冠状动脉标识数据的所述样本输入神经网络模型,对神经网络模型进行训练包括:将带有所述待识别血管扩充数据和所述冠状动脉启示数据的所述样本对神经网络模型进行训练,获得冠状动脉分割数据;基于所述冠状动脉分割数据与所述冠状动脉标识数据获得损失结果;以及基于所述损失结果,调整所述神经网络模型的参数。
- 根据权利要求5至7中任一所述的冠状动脉分割方法,其中,所述获取所述样本包括:获取原始图像,所述原始图像包含待识别血管和冠状动脉标识;以及对所述原始图像进行滑窗处理,获取多个所述样本。
- 一种冠状动脉分割装置,包括:分割获取模块,配置为获取待检测区域以及区域冠状动脉启示数据,所述区域冠状动脉启示数据用于在对所述待检测区域进行分割时提供启示;分割模块,配置为将带有所述区域冠状动脉启示数据的所述待检测区域输入到预先训练的冠状动脉分割模型,获取区域冠状动脉分割数据;比对判断模块,配置为获取所述区域冠状动脉启示数据与所述区域冠状动脉分割数据的比对结果;替换模块,配置为用所述比对结果替换待检测图像中的所述待检测区域,获取替换后的待检测图像,以完成检测步骤;输出模块,配置为重复所述检测步骤,其中,对于所述待检测图像对应的多个所述检测步骤,在上一个检测步骤中的所述替换后的待检测图像中选取下一个检测步骤的所述待检测区域,将上一个检测步骤中的所述比对结果的一部分作为下一个检测步骤中的所述冠状动脉启示数据;以及当所述待检测图像全部被所述比对结果遍历替换完,输出冠状动脉分割数据。
- 根据权利要求9所述的冠状动脉分割装置,其中,所述待检测区域为冠状动脉起源区域;其中,所述分割获取模块包括:起源分割子模块,配置为基于所述待检测图像,获取冠状动脉起源分割数据;第一提取子模块,配置为在所述冠状动脉起源分割数据中提取第一骨架;第一种子点选取子模块,配置为在所述第一骨架上按照第一预设步长选取第一种子点;第一栈顶种子点获取子模块,配置为将所述第一种子压入种子点栈中,获取第一栈顶种子点;第一区域获取子模块,配置为以所述第一栈顶种子点的中心为中心,在所述待检测图像上,选取第一预设尺寸大小的第一区域作为所述待检测区域;以及第一启示获取子模块,配置为将所述第一区域内的所述冠状动脉起源分割数据作为所述区域冠状动脉启示数据。
- 根据权利要求9或10所述的冠状动脉分割装置,其中,所述待检测区域来自于所述 上一个检测步骤中的替换后的待检测图像;其中,所述分割获取模块进一步包括:新增获取子模块,配置为在所述上一个检测步骤中的比对结果中获取新增冠状动脉分割数据;第二提取子模块,配置为在所述新增冠状动脉分割数据提取中第二骨架;第二种子点选取子模块,配置为在所述第二骨架上按照第二预设步长选取第二种子点;第二栈顶种子点获取子模块,配置为将所述第二种子点压入所述种子点栈中,获取第二栈顶种子点;第二区域获取子模块,配置为以所述第二栈顶种子点的中心为中心,在所述上一个检测步骤中的替换后的待检测图像上,选取第二预设尺寸大小的第二区域作为所述待检测区域,所述第二区域覆盖部分所述上一个检测步骤中的比对结果;第二启示获取子模块,配置为将所述第二区域内的所述上一个检测步骤中比对结果作为所述区域冠状动脉启示数据。
- 根据权利要求9至11中任一所述的冠状动脉分割装置,其中,所述比对判断模块进一步配置为当所述区域冠状动脉分割数据比所述区域冠状动脉启示数据分割出更多的分割区域,选取所述区域冠状动脉分割数据为所述比对结果;或当所述区域冠状动脉分割数据比所述区域冠状动脉启示数据并未分割出更多的分割区域,选取所述区域冠状动脉启示数据为所述比对结果。
- 根据权利要求9至12中任一所述的冠状动脉分割装置,其中,还包括:样本获取模块,配置为获取样本,所述样本包括待识别区域血管数据与冠状动脉标识数据;扩充模块,配置为在所述待识别区域血管数据中增加干扰数据,获取待识别血管扩充数据;启示获取模块,配置为基于所述冠状动脉标识数据获取冠状动脉启示数据,所述冠状动脉启示数据用于在对所述待识别区域血管数据进行标识时为神经网络模型提供一个准确的起源点提供启示,其中,所述基于所述冠状动脉标识数据获取冠状动脉启示数据,包括:从所述冠状动脉标识数据中选取部分作为所述冠状动脉启示数据;标识模块,配置为将带有所述待识别血管扩充数据、所述冠状动脉启示数据和所述冠状动脉标识数据的所述样本输入神经网络模型,对神经网络模型进行训练,以使得所述神经网络模型能够基于所述样本上的所述冠状动脉启示数据输出所述冠状动脉标识数据。
- 根据权利要求13所述的冠状动脉分割装置,其中,所述扩充模块进一步包括:噪声扩充子模块,配置为在所述待识别区域血管数据中添加噪声数据,所述噪声数据包括斑块数据和伪影数据;假阳扩充子模块,配置为在所述待识别区域血管数据中添加假阳数据。
- 根据权利要求13或14所述的冠状动脉分割装置,其中,所述标识模块进一步包括:输出子模块,配置为将带有所述待识别血管扩充数据和所述冠状动脉启示数据的所述样本对神经网络模型进行训练,获得冠状动脉分割数据;损失子模块,配置为基于所述冠状动脉分割数据与所述冠状动脉标识数据获得损失结果;调整子模块,配置为基于所述损失结果,调整所述神经网络模型的参数。
- 根据权利要求13至15中任一所述的所述的冠状动脉分割装置,其中,所述样本获取模块进一步包括:原始图像获取子模块,配置为获取原始图像,所述原始图像包含待识别血管和冠状动脉标识;以及预处理子模块,配置为对所述原始图像进行滑窗处理,获取多个所述样本。
- 一种电子设备,包括:处理器;以及存储器,在所述存储器中存储有计算机程序指令,所述计算机程序指令在被所述处理器运行时使得所述处理器执行如权利要求1至8中任一所述的冠状动脉分割方法。
- 一种计算机可读存储介质,所述计算机可读存储介质上存储有计算机程序指令,所述 计算机程序指令在被处理器运行时使得所述处理器执行如权利要求1至8中任一所述的冠状动脉分割方法。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2022518857A JP7349018B2 (ja) | 2020-11-19 | 2021-11-16 | 冠状動脈の分割方法、装置、電子デバイス及びコンピュータ読取可能な記憶媒体 |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011303653.9 | 2020-11-19 | ||
CN202011303653.9A CN112418299B (zh) | 2020-11-19 | 2020-11-19 | 冠状动脉分割模型训练方法、冠状动脉分割方法、及装置 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2022105735A1 true WO2022105735A1 (zh) | 2022-05-27 |
Family
ID=74773168
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2021/130861 WO2022105735A1 (zh) | 2020-11-19 | 2021-11-16 | 冠状动脉分割方法及其装置、电子设备及计算机可读存储介质 |
Country Status (3)
Country | Link |
---|---|
JP (1) | JP7349018B2 (zh) |
CN (1) | CN112418299B (zh) |
WO (1) | WO2022105735A1 (zh) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112418299B (zh) * | 2020-11-19 | 2021-10-26 | 推想医疗科技股份有限公司 | 冠状动脉分割模型训练方法、冠状动脉分割方法、及装置 |
CN112950734B (zh) * | 2021-03-01 | 2023-09-22 | 推想医疗科技股份有限公司 | 冠脉重建方法、装置、电子设备和存储介质 |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070118024A1 (en) * | 2005-11-23 | 2007-05-24 | General Electric | Iterative vascular identification |
US20110293150A1 (en) * | 2010-05-28 | 2011-12-01 | General Electric Company | Iterative vascular reconstruction by seed point segmentation |
US20130094725A1 (en) * | 2011-10-12 | 2013-04-18 | Siemens Corporation | Reproducible Segmentation of Elliptical Boundaries in Medical Imaging |
CN105427329A (zh) * | 2015-12-21 | 2016-03-23 | 东软集团股份有限公司 | 一种基于cta图像的双肢血管图像的分割方法和系统 |
CN110176013A (zh) * | 2019-05-29 | 2019-08-27 | 数坤(北京)网络科技有限公司 | 血管分割方法、装置、电子设备及存储介质 |
CN112418299A (zh) * | 2020-11-19 | 2021-02-26 | 推想医疗科技股份有限公司 | 冠状动脉分割模型训练方法、冠状动脉分割方法、及装置 |
Family Cites Families (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2007209589A (ja) * | 2006-02-10 | 2007-08-23 | Gifu Univ | 画像解析システム及び画像解析プログラム |
EP2220619A2 (en) * | 2007-12-14 | 2010-08-25 | Koninklijke Philips Electronics N.V. | Labeling a segmented object |
JP5388647B2 (ja) * | 2009-03-23 | 2014-01-15 | 株式会社東芝 | 医用画像処理装置及び方法、プログラム |
US10115039B2 (en) * | 2016-03-10 | 2018-10-30 | Siemens Healthcare Gmbh | Method and system for machine learning based classification of vascular branches |
CN108171698B (zh) * | 2018-02-12 | 2020-06-09 | 数坤(北京)网络科技有限公司 | 一种自动检测人体心脏冠脉钙化斑块的方法 |
CN109003278B (zh) * | 2018-03-26 | 2022-03-04 | 天津工业大学 | 一种改进的基于主动形状模型的ct图像主动脉分割方法 |
CN109272048B (zh) * | 2018-09-30 | 2022-04-12 | 北京工业大学 | 一种基于深度卷积神经网络的模式识别方法 |
CN110706233A (zh) * | 2019-09-30 | 2020-01-17 | 北京科技大学 | 一种视网膜眼底图像分割方法及装置 |
CN110706246B (zh) * | 2019-10-15 | 2022-08-02 | 上海微创卜算子医疗科技有限公司 | 一种血管图像分割方法、装置、电子设备和存储介质 |
CN111292314B (zh) * | 2020-03-03 | 2024-05-24 | 上海联影智能医疗科技有限公司 | 冠脉分段方法、装置、图像处理系统和存储介质 |
CN111724413B (zh) * | 2020-06-19 | 2021-09-28 | 上海杏脉信息科技有限公司 | 一种血管跟踪方法、冠脉影像分割方法、介质及电子设备 |
CN111815599B (zh) * | 2020-07-01 | 2023-12-15 | 上海联影智能医疗科技有限公司 | 一种图像处理方法、装置、设备及存储介质 |
-
2020
- 2020-11-19 CN CN202011303653.9A patent/CN112418299B/zh active Active
-
2021
- 2021-11-16 JP JP2022518857A patent/JP7349018B2/ja active Active
- 2021-11-16 WO PCT/CN2021/130861 patent/WO2022105735A1/zh active Application Filing
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070118024A1 (en) * | 2005-11-23 | 2007-05-24 | General Electric | Iterative vascular identification |
US20110293150A1 (en) * | 2010-05-28 | 2011-12-01 | General Electric Company | Iterative vascular reconstruction by seed point segmentation |
US20130094725A1 (en) * | 2011-10-12 | 2013-04-18 | Siemens Corporation | Reproducible Segmentation of Elliptical Boundaries in Medical Imaging |
CN105427329A (zh) * | 2015-12-21 | 2016-03-23 | 东软集团股份有限公司 | 一种基于cta图像的双肢血管图像的分割方法和系统 |
CN110176013A (zh) * | 2019-05-29 | 2019-08-27 | 数坤(北京)网络科技有限公司 | 血管分割方法、装置、电子设备及存储介质 |
CN112418299A (zh) * | 2020-11-19 | 2021-02-26 | 推想医疗科技股份有限公司 | 冠状动脉分割模型训练方法、冠状动脉分割方法、及装置 |
Also Published As
Publication number | Publication date |
---|---|
CN112418299B (zh) | 2021-10-26 |
CN112418299A (zh) | 2021-02-26 |
JP2023506107A (ja) | 2023-02-15 |
JP7349018B2 (ja) | 2023-09-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111899245B (zh) | 图像分割、模型的训练方法及装置,电子设备及存储介质 | |
Wu et al. | Predict pneumonia with chest X-ray images based on convolutional deep neural learning networks | |
Zhang et al. | Deep supervision with additional labels for retinal vessel segmentation task | |
Zuo et al. | R2AU-Net: attention recurrent residual convolutional neural network for multimodal medical image segmentation | |
CN109919928B (zh) | 医学影像的检测方法、装置和存储介质 | |
WO2022105735A1 (zh) | 冠状动脉分割方法及其装置、电子设备及计算机可读存储介质 | |
CN110475505A (zh) | 利用全卷积网络的自动分割 | |
US11972571B2 (en) | Method for image segmentation, method for training image segmentation model | |
AU2015207945A1 (en) | Method for training an artificial neural network | |
CN111899244B (zh) | 图像分割、网络模型的训练方法及装置,及电子设备 | |
Xie et al. | Optic disc and cup image segmentation utilizing contour-based transformation and sequence labeling networks | |
Barua et al. | An accurate valvular heart disorders detection model based on a new dual symmetric tree pattern using stethoscope sounds | |
CN112288718B (zh) | 图像处理方法及装置,电子设备及计算机可读存储介质 | |
Tajudin et al. | Deep learning in the grading of diabetic retinopathy: A review | |
Gordaliza et al. | Translational lung imaging analysis through disentangled representations | |
CN111445456A (zh) | 分类模型、网络模型的训练方法及装置、识别方法及装置 | |
Wang et al. | AVDNet: Joint coronary artery and vein segmentation with topological consistency | |
CN113011340B (zh) | 一种基于视网膜图像的心血管手术指标风险分类方法及系统 | |
CN114998582A (zh) | 一种冠脉血管的分割方法、装置及存储介质 | |
Sandhya et al. | Pixel‐Boundary‐Dependent Segmentation Method for Early Detection of Diabetic Retinopathy | |
CN111598870A (zh) | 基于卷积神经网络端对端推理计算冠状动脉钙化比的方法 | |
Munira et al. | Multi-Classification of Brain MRI Tumor Using ConVGXNet, ConResXNet, and ConIncXNet | |
Xu et al. | A Tuberculosis Detection Method Using Attention and Sparse R-CNN. | |
Lei et al. | GNN-fused CapsNet with multi-head prediction for diabetic retinopathy grading | |
Vinta et al. | Segmentation and Classification of Interstitial Lung Diseases Based on Hybrid Deep Learning Network Model |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
ENP | Entry into the national phase |
Ref document number: 2022518857 Country of ref document: JP Kind code of ref document: A |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 21893882 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
32PN | Ep: public notification in the ep bulletin as address of the adressee cannot be established |
Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 12.09.2023) |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 21893882 Country of ref document: EP Kind code of ref document: A1 |