CN113592764A - System and method for identifying and segmenting aneurysms of different anatomical regions of different modalities - Google Patents
System and method for identifying and segmenting aneurysms of different anatomical regions of different modalities Download PDFInfo
- Publication number
- CN113592764A CN113592764A CN202010371199.4A CN202010371199A CN113592764A CN 113592764 A CN113592764 A CN 113592764A CN 202010371199 A CN202010371199 A CN 202010371199A CN 113592764 A CN113592764 A CN 113592764A
- Authority
- CN
- China
- Prior art keywords
- aneurysm
- blood vessel
- dimensional surface
- surface model
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 206010002329 Aneurysm Diseases 0.000 title claims abstract description 105
- 238000000034 method Methods 0.000 title claims abstract description 27
- 210000003484 anatomy Anatomy 0.000 title claims abstract description 16
- 210000004204 blood vessel Anatomy 0.000 claims abstract description 108
- 238000002372 labelling Methods 0.000 claims abstract description 28
- 238000013135 deep learning Methods 0.000 claims abstract description 17
- 230000000877 morphologic effect Effects 0.000 claims abstract description 4
- 238000000605 extraction Methods 0.000 claims description 21
- 230000002792 vascular Effects 0.000 claims description 13
- 230000011218 segmentation Effects 0.000 claims description 12
- 238000009499 grossing Methods 0.000 claims description 7
- 238000004891 communication Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000005316 response function Methods 0.000 claims description 5
- 210000001367 artery Anatomy 0.000 claims description 4
- 210000001715 carotid artery Anatomy 0.000 claims description 3
- 238000012937 correction Methods 0.000 claims description 3
- 230000006870 function Effects 0.000 claims description 3
- 238000007917 intracranial administration Methods 0.000 claims description 3
- 230000002708 enhancing effect Effects 0.000 claims description 2
- 208000007474 aortic aneurysm Diseases 0.000 description 5
- 201000008450 Intracranial aneurysm Diseases 0.000 description 4
- 201000008982 Thoracic Aortic Aneurysm Diseases 0.000 description 3
- 208000003457 familial thoracic 1 aortic aneurysm Diseases 0.000 description 3
- 208000002223 abdominal aortic aneurysm Diseases 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 230000003187 abdominal effect Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000002583 angiography Methods 0.000 description 1
- 210000000709 aorta Anatomy 0.000 description 1
- 210000000702 aorta abdominal Anatomy 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000005728 strengthening Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000008961 swelling Effects 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/08—Projecting images onto non-planar surfaces, e.g. geodetic screens
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background 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
- G06T2207/30096—Tumor; Lesion
-
- 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)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Apparatus For Radiation Diagnosis (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention belongs to the technical field of biomedicine, relates to a medical image application system and method, and particularly relates to a system and method capable of identifying and segmenting aneurysms of different anatomical parts in different modes. The method comprises a model training part or a model prediction part, wherein the aneurysm is identified or segmented by respectively modeling and comparing a target blood vessel segment with a normal blood vessel three-dimensional surface. The invention can extract and reserve the common morphological characteristics of all aneurysms, such as the limitation of the wall structure or the diffuse expansion and bulging, so that the application of deep learning becomes possible; the workload and difficulty of aneurysm data labeling can be remarkably reduced, the quality and quantity of labeled data are increased, and the generalization capability of a deep learning scheme is enhanced.
Description
Technical Field
The invention belongs to the field of biomedicine, relates to a medical image application system and method, and particularly relates to a system and method capable of identifying and segmenting aneurysms of different anatomical parts in different modes.
Background
In recent years, deep learning has gained wide attention and application in the field of medical images, and has been increasingly used for identification of aneurysms. However, the current deep learning scheme still has the following limitations: 1) the training model is limited to a single modality, a single site. For example, MRA image-based intracranial aneurysm recognition [1], CTA-based intracranial aneurysm recognition [2], CT-based abdominal aortic aneurysm recognition [3], and the like, the differences in the expression of image data based on gray information on different anatomical parts and different modalities are too large, so that the generalization capability of a deep learning model is reduced [4], 2) the amount of training data is small. The data volume is generally in the order of hundreds of orders of magnitude based on deep learning models of medical images [1-3 ]. The main reason is that the labeling of the image data needs to consume a lot of time and energy, resulting in less high-quality labeled data, thereby indirectly affecting the generalization ability of the model.
Based on the current state of the art, the inventors of the present application propose to provide a system and method that can identify and segment aneurysms of different anatomical regions of different modalities.
The prior art related to the present invention is:
[1]Dai ju Ueda,Akira Yamamoto,Masataka Nishimoriet,et al.(2018).Deep Learning for MR Angiography:Automated Detection of Cerebral Aneurysms.Radiology.290.180901.
[2]Allison Park,Chris Chute,Pranav Rajpurkar,et,al.(2019).Deep Learning-Assisted Diagnosis of Cerebral Aneurysms Using the HeadXNet Model.JAMA network open.2.e195600.10.1001/jamanetworkopen.2019.5600.
[3]Jen-Tang Lu,Rupert Brooks,Stefan Hahn,et,al.(2019).DeepAAA:clinically applicable and generalizable detection of abdominal aortic aneurysm using deep learning.10.1007/978-3-030-32245-8_80.
[4]John Zech,Marcus Badgeley,Manway Liu,et al.(2018).Variable generalization performance of a deep learning model to detect pneumonia in chest radiographs:A cross-sectional study.PLOS Medicine.15.e1002683.10.1371/journal.pmed.1002683.。
disclosure of Invention
It is an object of the present invention to provide a system and method for identifying and segmenting aneurysms of different anatomical regions of different modalities based on the prior art basis that aneurysms can occur in various regions of the human arterial system. The invention can extract and retain the common morphological characteristics (the limitation or diffuse expansion and swelling of the wall structure) of all the aneurysms, so that the application of deep learning becomes possible; meanwhile, the workload and difficulty of aneurysm data labeling can be remarkably reduced, so that the quality and quantity of labeled data are increased, and the generalization capability of a deep learning scheme is enhanced.
Specifically, the invention provides a system for identifying and segmenting aneurysms of different anatomical regions in different modalities, which comprises a model training part or a model part, wherein the aneurysms are identified or segmented by respectively modeling and comparing a target blood vessel section and a normal blood vessel three-dimensional surface.
The model training part comprises the following modules:
the blood vessel extraction module is used for extracting blood vessels by a self-adaptive threshold value and region growing method;
the blood vessel three-dimensional surface model generating module is used for converting the segmented binary image into a three-dimensional surface model by adopting a marching cubes algorithm and smoothing the three-dimensional surface model so as to remove noise and a non-manifold surface;
an aneurysm labeling module for labeling by using a three-dimensional surface model;
the aneurysm vascular section extraction module selects a vascular range with a geodesic distance from the marked aneurysm smaller than a certain threshold value as a final vascular section;
the normal blood vessel section extraction module randomly selects a blood vessel range with the geodesic distance smaller than a certain threshold value as a normal blood vessel section by taking the blood vessel intersection as a center;
and the model training module is used for visually displaying the three-dimensional topological structure and the form of the aneurysm by using different convolution structures for the three-dimensional surface model.
Preferably, for the human aortic blood vessels, the blood vessel extraction is realized by the following steps:
enhancement of blood vessels: and carrying out calculation based on the multi-scale Hessian characteristic value on the image. Any one of Frangi, Sato and Jerman characteristic response functions can be adopted;
estimating the center line and the radius of the blood vessel: based on the blood vessel enhanced image, carrying out Dijkstra shortest path search to obtain a central line, wherein the radius is a scale value corresponding to the maximum characteristic response function value;
vessel segmentation: classifying the foreground and the background of the original image according to the central line and the radius, and then segmenting by adopting a graph cut: the user can make appropriate corrections to the final vessel segmentation result.
Preferably, compared with a layer-by-layer labeling mode based on an original image, the aneurysm labeling module only modifies the aneurysm and the blood vessel region every time of labeling, and does not have a background region covering the blood vessel and the aneurysm; and the range of the aneurysm is confirmed without multiple contrasts of adjacent layers during labeling.
Preferably, the normal blood vessel extraction module is used for ensuring the balance of positive and negative samples in the training data set, and the number of the selected normal blood vessel sections is consistent with that of the aneurysm blood vessel sections.
Preferably, the model prediction part comprises:
the blood vessel extraction module is used for extracting blood vessels by a self-adaptive threshold value and region growing method;
the blood vessel three-dimensional surface model generating module is used for converting the segmented binary image into a three-dimensional surface model by adopting a marching cubes algorithm and smoothing the three-dimensional surface model so as to remove noise and a non-manifold surface;
a suspected aneurysm positioning module for positioning aneurysm in an automatic or semi-automatic manner;
the target blood vessel section extraction module selects a blood vessel range with a distance to the seed point measuring ground smaller than a certain threshold value as a target blood vessel section;
the aneurysm segmentation module is used for loading the trained model and carrying out deep learning prediction on the extracted target blood vessel section to obtain a segmentation result of the aneurysm;
and the aneurysm display module is used for displaying the segmented aneurysm and the three-dimensional surface model of the blood vessel in the original image in an overlapping mode.
Wherein, the automatic mode is as follows:
carrying out multi-scale Hessian characteristic value calculation on a blood vessel image region in a blood vessel extraction module, and extracting regions of which the characteristic values are all negative numbers and the corresponding function values of the characteristics are higher than a certain threshold value;
performing communication domain analysis to obtain a plurality of communication areas;
and taking the centroid position of each connected region as a seed point.
The semi-automatic mode is that a suspected aneurysm position is manually selected by a user based on a three-dimensional blood vessel surface model generated by a blood vessel three-dimensional surface model generation module and marked by a seed point.
Accordingly, the present invention provides a method of identifying and segmenting aneurysms. The method comprises model training or model prediction;
the model training comprises the following steps:
blood vessel extraction, for small artery blood vessels of human body such as carotid artery, intracranial artery, etc. The extraction can be carried out by a self-adaptive threshold value and a region growing method;
generating a blood vessel three-dimensional surface model, converting the segmented binary image into a three-dimensional surface model by adopting a marching cubes algorithm, and smoothing the three-dimensional surface model to remove noise and a non-manifold surface;
labeling aneurysms, namely labeling by using a three-dimensional surface model, and completing the labeling work of a single aneurysm by a user only needing several strokes;
extracting an aneurysm vascular segment, and selecting a vascular range with a geodesic distance from the marked aneurysm smaller than a certain threshold value as a final vascular segment;
extracting a normal blood vessel section, wherein a blood vessel range with the geodesic distance smaller than a certain threshold value is randomly selected as the normal blood vessel section by taking the blood vessel intersection as the center; in order to ensure the balance of positive and negative samples in the training data set, the number of the selected normal blood vessel sections is consistent with that of the aneurysm blood vessel sections;
model training, wherein the three-dimensional surface model can visually display the three-dimensional topological structure and the form of the aneurysm, but has a larger difference with a three-dimensional gray image and needs to use different convolution structures;
the model prediction comprises the following steps:
blood vessel extraction;
generating a three-dimensional surface model of the blood vessel;
suspected aneurysm positioning, namely positioning the aneurysm in an automatic or semi-automatic mode;
extracting a target blood vessel section, and selecting a blood vessel range with a seed point measuring distance smaller than a certain threshold value in the step of positioning the suspected aneurysm as the target blood vessel section;
dividing the aneurysm, loading a trained model, and performing deep learning prediction on the blood vessel section extracted in the suspected aneurysm positioning step to obtain a dividing result of the aneurysm;
displaying the aneurysm, namely displaying the segmented aneurysm in an overlapping way with the original image or the blood vessel three-dimensional surface model in the blood vessel three-dimensional surface model generating step.
The system and method of the present invention can be used to extract and retain morphological features common to all aneurysms, enabling the application of deep learning, or even enhancing the generalization ability of deep learning schemes.
The invention has the advantages that:
1) the method can be applied to the aneurysm identification and segmentation tasks of a plurality of modalities and a plurality of anatomical parts. Such as a carotid or intracranial aneurysm based on CTA, MRA, 3D-DSA modalities, an abdominal or thoracic aortic aneurysm based on CTA, MRA, and so forth.
2) And the data marking is simpler and faster.
3) The training model is small in size and easy to converge. Mobile-end deployment may be implemented.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
Fig. 1. aneurysm automatic identification process.
FIG. 2. model prediction module.
FIG. 3 is an example of a rapid annotation of a thoracic aortic aneurysm. The aneurysm labeling can be completed only by several strokes, and the labeling brush is limited in the three-dimensional surface model area without affecting the background area. The screenshots were derived from the open source software MeshLab.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
The invention is based on a deep learning scheme and consists of two independent functional modules of training and prediction, and the flow charts are respectively shown in fig. 1 and fig. 2.
Example 1 model training Module part
The method comprises the following steps:
1) and (5) extracting blood vessels. For small arterial vessels of the human body such as carotid arteries, intracranial arteries, etc. Extraction can be performed by adaptive threshold and region growing methods. For the blood vessels of the aorta of the human body, such as the abdominal aorta and the like, the method can be realized by the following steps: a) and (4) strengthening blood vessels. And carrying out calculation based on the multi-scale Hessian characteristic value on the image. Any of Frangi, Sato, and Jerman characteristic response functions may be used. b) Vessel centerline and radius estimation. And based on the blood vessel enhanced image, carrying out Dijkstra shortest path search to obtain a central line. The radius is the scale value corresponding to the maximum characteristic response function value. c) And (5) segmenting blood vessels. And carrying out foreground and background classification on the original image according to the central line and the radius, and then segmenting by adopting a graph cut. The user can make appropriate corrections to the final vessel segmentation result.
2) And generating a three-dimensional surface model of the blood vessel. And converting the segmented binary image into a three-dimensional surface model by using a marching cubes algorithm, and smoothing the three-dimensional surface model to remove noise and a non-manifold (non-manifold) surface.
3) And (5) labeling the aneurysm. By using the three-dimensional surface model for labeling, the user can finish the labeling work of a single aneurysm only by strokes. Compared with the mode of layer-by-layer labeling based on the original image, the method has the obvious advantages that a) only the aneurysm and the blood vessel area can be modified during each labeling, and the background area which covers the blood vessel and the aneurysm does not exist; b) the aneurysm display is more intuitive, and the range of the aneurysm is confirmed without multiple contrasts of adjacent layers during marking. Fig. 3 is an example of rapid labeling of a thoracic aortic aneurysm using an open source tool.
4) And (5) extracting an aneurysm vascular section. And selecting a blood vessel range with the geodesic distance (geodesic distance) to the marked aneurysm smaller than a certain threshold value as a final blood vessel section.
5) And (4) extracting a normal blood vessel section. And randomly selecting a blood vessel range with the geodesic distance smaller than a certain threshold value as a normal blood vessel section by taking the blood vessel intersection point as a center. To ensure that the positive and negative samples are balanced in the training data set, the selected normal vessel segments should be consistent in number with the aneurysm vessel segments.
6) And (5) training a model. The three-dimensional surface model can visually display the three-dimensional topological structure and the form of the aneurysm, but has a larger difference with a three-dimensional gray image, and different convolution structures such as MeshCNN and the like need to be used. Because the training set is derived from the extracted three-dimensional surface model, the defects of overlarge data size and insufficient GPU memory required by three-dimensional model training are effectively overcome, and therefore, alternative schemes of sacrificing model accuracy, such as down-sampling images, random blocking of three-dimensional images or 2D layer-by-layer training and the like, are avoided. Similarly, the model training parameters are less, the model volume is greatly compressed and can be controlled within 10M, and the deployment of a mobile terminal is facilitated.
Example 2 model prediction Module part
The method comprises the following steps:
1) and (5) extracting blood vessels. As above.
2) And generating a three-dimensional surface model of the blood vessel. As above.
3) Suspected aneurysm localization. The aneurysm may be located in an automated or semi-automated manner. The automatic mode comprises the following steps: a) carrying out multi-scale Hessian characteristic value calculation on the blood vessel image region in the step 1), and extracting regions of which the characteristic values are all negative numbers and the corresponding function values of the characteristics are higher than a certain threshold value; b) performing communication domain analysis to obtain a plurality of communication areas; c) and taking the centroid position of each connected region as a seed point. The semi-automatic mode is as follows: b) based on the three-dimensional blood vessel surface model generated in the step 2), the suspected aneurysm position is manually selected by a user and marked by a seed point.
4) And extracting a target blood vessel section. And 3) selecting a blood vessel range with the geodesic distance from the seed point in the step 3) being less than a certain threshold value as a target blood vessel section.
5) And (5) dividing the aneurysm. Loading a trained model, and carrying out deep learning prediction on the blood vessel section extracted in the step 4) to obtain a segmentation result of the aneurysm.
6) Displaying the aneurysm. Displaying the segmented aneurysm in superposition with the original image or the blood vessel three-dimensional surface model in the step 2).
The above description is only for the specific embodiments of the present application, but the scope of the present application is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present disclosure should be covered within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (10)
1. A system for identifying and segmenting aneurysms of different anatomical regions of different modalities, the system comprising a model training or prediction component that identifies or segments the aneurysm by separately modeling and comparing a target vessel segment to a three-dimensional surface of a normal vessel.
2. The system for identifying and segmenting aneurysms of different anatomical regions of different modalities according to claim 1, wherein the model training component comprises the following modules:
the blood vessel extraction module is used for extracting blood vessels by a self-adaptive threshold value and region growing method;
the blood vessel three-dimensional surface model generation module is used for converting the segmented binary image into a three-dimensional surface model and smoothing the three-dimensional surface model so as to remove noise and a non-manifold surface;
an aneurysm labeling module for labeling by using a three-dimensional surface model;
the aneurysm vascular section extraction module selects a vascular range with a geodesic distance from the marked aneurysm smaller than a certain threshold value as a final vascular section;
the normal blood vessel section extraction module randomly selects a blood vessel range with the geodesic distance smaller than a certain threshold value as a normal blood vessel section by taking the blood vessel intersection as a center;
and the model training module is used for visually displaying the three-dimensional topological structure and the form of the aneurysm by using different convolution structures for the three-dimensional surface model.
3. The system for identifying and segmenting aneurysms of different anatomical regions of different modalities according to claim 2, wherein the system is implemented for human aortic vessels, vessel extraction by:
a) enhancement of blood vessels: calculating the image based on a multi-scale Hessian characteristic value;
b) estimating the center line and the radius of the blood vessel: based on the blood vessel enhanced image, carrying out Dijkstra shortest path search to obtain a central line, wherein the radius is a scale value corresponding to the maximum characteristic response function value;
c) vessel segmentation: classifying the foreground and the background of the original image according to the central line and the radius, and then segmenting by adopting a graph cut: the user can make appropriate corrections to the final vessel segmentation result.
4. The system for identifying and segmenting aneurysms of different anatomical regions of different modalities according to claim 2, wherein the aneurysm labeling module modifies only the aneurysm and the vessel region per labeling without having background regions that cover the vessels and the aneurysm; and the range of the aneurysm is confirmed without multiple contrasts of adjacent layers during labeling.
5. The system according to claim 2, wherein the normal vessel extraction module selects normal vessel segments consistent in number with the aneurysm vessel segments to ensure positive and negative sample balance in the training data set.
6. The system for identifying and segmenting aneurysms of different anatomical regions of a modality according to claim 1, wherein the model prediction component comprises:
the blood vessel extraction module is used for extracting blood vessels by a self-adaptive threshold value and region growing method;
the blood vessel three-dimensional surface model generating module is used for converting the segmented binary image into a three-dimensional surface model by adopting a marching cubes algorithm and smoothing the three-dimensional surface model so as to remove noise and a non-manifold surface;
a suspected aneurysm positioning module for positioning aneurysm in an automatic or semi-automatic manner;
the target blood vessel section extraction module selects a blood vessel range with a distance to the seed point measuring ground smaller than a certain threshold value as a target blood vessel section;
the aneurysm segmentation module is used for loading the trained model and carrying out deep learning prediction on the extracted target blood vessel section to obtain a segmentation result of the aneurysm;
and the aneurysm display module is used for displaying the segmented aneurysm and the three-dimensional surface model of the blood vessel in the original image in an overlapping mode.
7. The system for identifying and segmenting aneurysms of different anatomical regions of a modality according to claim 6, wherein the automated means is:
a) carrying out multi-scale Hessian characteristic value calculation on a blood vessel image region in a blood vessel extraction module, and extracting regions of which the characteristic values are all negative numbers and the corresponding function values of the characteristics are higher than a certain threshold value;
b) performing communication domain analysis to obtain a plurality of communication areas;
c) and taking the centroid position of each connected region as a seed point.
8. The system according to claim 6, wherein the semi-automatic means is to manually select the location of the suspected aneurysm by a user and mark the location with a seed point based on the three-dimensional vessel surface model generated by the vessel three-dimensional surface model generation module.
9. A method for identifying and segmenting aneurysms based on the claimed system for identifying and segmenting aneurysms of different anatomical regions of different modalities, the method comprising model training or model prediction;
the model training comprises the following steps:
A) extracting blood vessels, namely extracting small arterial blood vessels of a human body such as carotid artery and intracranial artery by a self-adaptive threshold value and region growing method;
B) generating a blood vessel three-dimensional surface model, converting the segmented binary image into a three-dimensional surface model by adopting a marching cubes algorithm, and smoothing the three-dimensional surface model to remove noise and a non-manifold surface;
C) labeling aneurysms, namely labeling by using a three-dimensional surface model, and completing the labeling work of a single aneurysm by a user only needing several strokes;
D) extracting an aneurysm vascular segment, and selecting a vascular range with a geodesic distance from the marked aneurysm smaller than a certain threshold value as a final vascular segment;
E) extracting a normal blood vessel section, wherein a blood vessel range with the geodesic distance smaller than a certain threshold value is randomly selected as the normal blood vessel section by taking the blood vessel intersection as the center; the number of the selected normal blood vessel sections is consistent with that of the aneurysm blood vessel sections so as to ensure the balance of positive and negative samples in the training data set;
F) model training, wherein a three-dimensional surface model visually displays the three-dimensional topological structure and the form of the aneurysm, but the three-dimensional surface model is greatly different from a three-dimensional gray image and needs to use different convolution structures;
the model prediction comprises the following steps:
1) blood vessel extraction;
2) generating a three-dimensional surface model of the blood vessel;
3) suspected aneurysm positioning, namely positioning the aneurysm in an automatic or semi-automatic mode;
4) extracting a target blood vessel section, namely selecting a blood vessel range with the geodesic distance from the seed point in the step 3) being less than a certain threshold value as the target blood vessel section;
5) segmenting the aneurysm, loading a trained model, and performing deep learning prediction on the blood vessel section extracted in the step 4) to obtain an aneurysm segmentation result;
6) displaying the aneurysm, and displaying the segmented aneurysm in superposition with the original image or the blood vessel three-dimensional surface model in the step 2).
10. Use of the system of any one of claims 1 to 8 for extracting and preserving morphological features common to all aneurysms, enabling the application of deep learning, enhancing the generalization ability of deep learning schemes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010371199.4A CN113592764A (en) | 2020-05-01 | 2020-05-01 | System and method for identifying and segmenting aneurysms of different anatomical regions of different modalities |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010371199.4A CN113592764A (en) | 2020-05-01 | 2020-05-01 | System and method for identifying and segmenting aneurysms of different anatomical regions of different modalities |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113592764A true CN113592764A (en) | 2021-11-02 |
Family
ID=78238026
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010371199.4A Pending CN113592764A (en) | 2020-05-01 | 2020-05-01 | System and method for identifying and segmenting aneurysms of different anatomical regions of different modalities |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113592764A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116503395A (en) * | 2023-06-26 | 2023-07-28 | 杭州脉流科技有限公司 | Method, device and equipment for automatically obtaining morphological parameters aiming at wide-neck aneurysm |
| CN116863146A (en) * | 2023-06-09 | 2023-10-10 | 强联智创（北京）科技有限公司 | Method, apparatus and storage medium for extracting hemangio features |
-
2020
- 2020-05-01 CN CN202010371199.4A patent/CN113592764A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116863146A (en) * | 2023-06-09 | 2023-10-10 | 强联智创（北京）科技有限公司 | Method, apparatus and storage medium for extracting hemangio features |
| CN116863146B (en) * | 2023-06-09 | 2024-03-08 | 强联智创（北京）科技有限公司 | Method, apparatus and storage medium for extracting hemangio features |
| CN116503395A (en) * | 2023-06-26 | 2023-07-28 | 杭州脉流科技有限公司 | Method, device and equipment for automatically obtaining morphological parameters aiming at wide-neck aneurysm |
| CN116503395B (en) * | 2023-06-26 | 2023-09-08 | 杭州脉流科技有限公司 | Method, device and equipment for automatically obtaining morphological parameters aiming at wide-neck aneurysm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110706246B (en) | Blood vessel image segmentation method and device, electronic equipment and storage medium | |
| CN111696089B (en) | Arteriovenous determination method, device, equipment and storage medium | |
| EP1851722B1 (en) | Image processing device and method | |
| CN108280827B (en) | Coronary artery lesion automatic detection method, system and equipment based on deep learning | |
| JP4914517B2 (en) | Structure detection apparatus and method, and program | |
| US9888896B2 (en) | Determining a three-dimensional model dataset of a blood vessel system with at least one vessel segment | |
| EP1238369A1 (en) | Knowledge based computer aided diagnosis | |
| US9730609B2 (en) | Method and system for aortic valve calcification evaluation | |
| US20200410685A1 (en) | Method and system for acquiring status of strain and stress of a vessel wall | |
| CN109035194B (en) | Blood vessel extraction method and device | |
| CN112308846B (en) | Blood vessel segmentation method and device and electronic equipment | |
| CN111028248A (en) | Method and device for separating static and dynamic pulses based on CT (computed tomography) image | |
| US8295569B2 (en) | Method and system for automatic detection and measurement of mitral valve inflow patterns in doppler echocardiography | |
| CN113592764A (en) | System and method for identifying and segmenting aneurysms of different anatomical regions of different modalities | |
| CN112862835A (en) | Coronary vessel segmentation method, device, equipment and computer readable storage medium | |
| CN110163872A (en) | A kind of method and electronic equipment of HRMR image segmentation and three-dimensional reconstruction | |
| CN115953393B (en) | Intracranial aneurysm detection system, device and storage medium based on multitask learning | |
| CN112991365A (en) | Coronary artery segmentation method, system and storage medium | |
| Blondel et al. | Automatic trinocular 3D reconstruction of coronary artery centerlines from rotational X-ray angiography | |
| Zhang et al. | Attention-guided feature extraction and multiscale feature fusion 3d resnet for automated pulmonary nodule detection | |
| JP5395868B2 (en) | Image processing apparatus and method, and program | |
| Cui et al. | Coronary artery segmentation via hessian filter and curve-skeleton extraction | |
| CN110689080A (en) | Planar atlas construction method of blood vessel structure image | |
| CN116563305A (en) | Segmentation method and device for abnormal region of blood vessel and electronic equipment | |
| CN112862785B (en) | CTA image data identification method, device and storage medium |
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 |