CN116563305A - Segmentation method and device for abnormal region of blood vessel and electronic equipment - Google Patents
Segmentation method and device for abnormal region of blood vessel and electronic equipment Download PDFInfo
- Publication number
- CN116563305A CN116563305A CN202310347024.3A CN202310347024A CN116563305A CN 116563305 A CN116563305 A CN 116563305A CN 202310347024 A CN202310347024 A CN 202310347024A CN 116563305 A CN116563305 A CN 116563305A
- Authority
- CN
- China
- Prior art keywords
- image
- blood vessel
- region
- point cloud
- abnormal
- 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
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 270
- 230000011218 segmentation Effects 0.000 title claims abstract description 240
- 210000004204 blood vessel Anatomy 0.000 title claims abstract description 226
- 238000000034 method Methods 0.000 title claims abstract description 74
- 230000000903 blocking effect Effects 0.000 claims abstract description 14
- 238000002372 labelling Methods 0.000 claims description 23
- 238000012545 processing Methods 0.000 claims description 14
- 230000006870 function Effects 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 7
- 238000004590 computer program Methods 0.000 claims description 6
- 230000000877 morphologic effect Effects 0.000 claims description 5
- 238000012805 post-processing Methods 0.000 claims description 3
- 206010002329 Aneurysm Diseases 0.000 description 29
- 230000008569 process Effects 0.000 description 19
- 238000010586 diagram Methods 0.000 description 7
- 230000002792 vascular Effects 0.000 description 7
- 238000006243 chemical reaction Methods 0.000 description 6
- 230000008901 benefit Effects 0.000 description 5
- 201000008450 Intracranial aneurysm Diseases 0.000 description 4
- 230000004888 barrier function Effects 0.000 description 4
- 238000001514 detection method Methods 0.000 description 4
- 238000000605 extraction Methods 0.000 description 4
- 238000003709 image segmentation Methods 0.000 description 4
- 238000012986 modification Methods 0.000 description 4
- 230000004048 modification Effects 0.000 description 4
- 238000005457 optimization Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000003384 imaging method Methods 0.000 description 3
- 208000012661 Dyskinesia Diseases 0.000 description 2
- 208000032851 Subarachnoid Hemorrhage Diseases 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000006798 recombination Effects 0.000 description 2
- 238000005215 recombination Methods 0.000 description 2
- 210000003484 anatomy Anatomy 0.000 description 1
- 238000002583 angiography Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/70—Labelling scene content, e.g. deriving syntactic or semantic representations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
-
- 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
- G06T2207/30016—Brain
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The application discloses a segmentation method and device for abnormal regions of blood vessels and electronic equipment, wherein the method comprises the following steps: acquiring a blood vessel image to be segmented, wherein the blood vessel image comprises a blood vessel region and an abnormal region attached to the blood vessel; acquiring a point cloud domain image comprising a blood vessel region and an abnormal region based on the blood vessel image; inputting the point cloud domain image into a pre-trained target segmentation model to acquire a segmentation image of an abnormal region, wherein the target segmentation model comprises a first segmentation module for segmenting the abnormal region and a second segmentation module for blocking a region adjacent to a boundary between a blood vessel region and the abnormal region. According to the method and the device, the point cloud domain image is segmented by the first segmentation module to obtain the segmented image of the initial abnormal region, the boundary between the blood vessel region and the initial abnormal region is determined, and the point cloud near the boundary is further blocked by the second segmentation module conveniently, so that the segmented image of the final abnormal region is accurately obtained.
Description
Technical Field
The present invention relates to the field of image segmentation technologies, and in particular, to a method and an apparatus for segmenting an abnormal region of a blood vessel, and an electronic device.
Background
Intracranial aneurysms (intracranial aneurysm, abbreviated as IA) are a common disease threatening the health of people, and have a ratio of about 3.2% -6% in the general population and about 80% -85% in patients with spontaneous subarachnoid hemorrhage (subarachnoid hemorrhage, abbreviated as SAH). Rupture of an aneurysm can lead to significantly high mortality, and even survivors can suffer from prolonged physical and psychological stress, resulting in a serious decline in quality of life.
With advances in technology and medical level, advanced imaging techniques are widely used, and aneurysms are increasingly discovered. In the existing aneurysm segmentation method, a segmentation method based on image domain data or a point cloud domain detection method is generally adopted. However, the segmentation method based on the image domain data not only needs to consume a large amount of computing resources, but also does not fully utilize the anatomical structure characteristics of the aneurysm, and the accuracy of the detection result is low; the segmentation method based on the point cloud domain also has the problem that the segmentation result is not accurate enough.
Therefore, a method for dividing abnormal regions of blood vessels is needed to solve the problem that the abnormal regions of blood vessels are not divided accurately in the prior art.
Disclosure of Invention
In view of the above, the present invention provides a method, an apparatus and an electronic device for dividing abnormal regions of blood vessels, which mainly aims to solve the problem that the current division of abnormal regions of blood vessels is not accurate enough.
In order to solve the above problem, the present application provides a segmentation method of an abnormal region of a blood vessel, including:
acquiring a blood vessel image to be segmented, wherein the blood vessel image comprises a blood vessel region and an abnormal region attached to the blood vessel;
acquiring a point cloud domain image containing the blood vessel region and the abnormal region based on the blood vessel image;
inputting the point cloud domain image into a pre-trained target segmentation model to acquire a segmentation image of the abnormal region, wherein the target segmentation model comprises a first segmentation module for segmenting the abnormal region and a second segmentation module for blocking a region adjacent to a boundary between the blood vessel region and the abnormal region.
Optionally, the acquiring, based on the blood vessel image, a point cloud domain image including the blood vessel region and the abnormal region includes:
Carrying out gridding treatment on blood vessels and abnormal areas in the blood vessel image, and extracting at least one pixel point in the grid to obtain a pixel point set;
converting the pixel point set into a point cloud domain, and acquiring a point cloud domain image; or alternatively, the first and second heat exchangers may be,
extracting boundary point pixels of a blood vessel region and an abnormal region in the blood vessel image to obtain a pixel point set;
and converting the pixel point set into a point cloud domain, and acquiring a point cloud domain image.
Optionally, the inputting the point cloud domain image into a pre-trained target segmentation model, and obtaining the segmented image of the abnormal region includes:
acquiring an initial segmentation image of the abnormal region and a boundary between the abnormal region and the blood vessel region based on the first segmentation module;
and performing blocking processing on point clouds in specific areas on two sides of a boundary between an abnormal area and a blood vessel area in the initial segmented image based on the second segmentation module, and acquiring segmented images of the abnormal area, wherein the blocking processing comprises classifying the point clouds on one side of the abnormal area as the abnormal area and classifying the point clouds on one side of the blood vessel area as the blood vessel area.
Optionally, before inputting the point cloud domain image into a pre-trained target segmentation model to obtain the segmented image of the abnormal region, the method further includes:
Dividing the point cloud domain image into a plurality of blood vessel segment images;
inputting the point cloud domain image into a pre-trained target segmentation model to obtain a segmented image of the abnormal region, wherein the method comprises the following steps:
inputting the plurality of blood vessel segment images into a pre-trained target segmentation model to obtain a plurality of sub-segmentation images of an abnormal region corresponding to the plurality of blood vessel segments;
and recombining the sub-divided images to obtain the divided images of the abnormal region.
Optionally, the method for segmenting the abnormal region of the blood vessel further includes: obtaining a pre-trained target segmentation model, wherein the obtaining the pre-trained target segmentation model comprises:
acquiring a sample point cloud domain image containing an abnormal region and a blood vessel and a label point cloud domain image marking the abnormal region;
training an initial first segmentation module based on the sample point cloud domain image and the label point cloud domain image, and acquiring the first segmentation module when a first preset termination condition is reached;
acquiring the boundary between the abnormal region and the blood vessel based on the label point cloud domain image;
and training an initial second segmentation module based on the boundary and tag point cloud domain image and the output image of the first segmentation module, and acquiring the second segmentation module when a second preset termination condition is reached, wherein the region of interest of the second segmentation module at least comprises regions in specific ranges on two sides of the boundary, the first preset termination condition comprises that the value of a loss function corresponding to the first segmentation module reaches a preset threshold value or the training times of the first segmentation module reaches preset times, and the second preset termination condition comprises that the value of the loss function corresponding to the second segmentation module reaches the preset threshold value or the training times of the second segmentation module reaches preset times.
Optionally, the segmentation method of the abnormal region of the blood vessel further includes: and determining the region of interest of the second segmentation module through a neighbor algorithm or a fixed threshold method.
Optionally, the obtaining a tag point cloud domain image for labeling the abnormal region includes:
acquiring the boundary between a blood vessel and an abnormal region in the sample point cloud domain image and a seed point in the abnormal region;
and labeling the abnormal region in the sample point cloud domain image based on the anatomical morphological characteristics, the boundary and the seed points of the abnormal region, and obtaining a label point cloud domain image.
Optionally, after obtaining the segmented image of the abnormal region, the method further comprises: and displaying the abnormal region and the blood vessel region in the point cloud domain image by adopting different display modes.
Optionally, the target segmentation model further includes a post-processing module, configured to restore an output result of the second segmentation module to an image domain.
In order to solve the above-mentioned problem, the present application provides a segmentation apparatus for an abnormal region of a blood vessel, comprising:
the first acquisition module is used for acquiring a blood vessel image to be segmented, wherein the blood vessel image comprises a blood vessel region and an abnormal region attached to the blood vessel;
The second acquisition module is used for acquiring a point cloud domain image containing the blood vessel region and the abnormal region based on the blood vessel image;
the segmentation module is used for inputting the point cloud domain image into a pre-trained target segmentation model to acquire a segmentation image of the abnormal region, wherein the target segmentation model comprises a first segmentation module used for segmenting the abnormal region and a second segmentation module used for blocking a region adjacent to the boundary between the blood vessel region and the abnormal region.
In order to solve the above problems, the present application provides an electronic device, at least including a memory, and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method for dividing the abnormal region of the blood vessel according to any one of the above steps when executing the computer program on the memory.
According to the segmentation method, the segmentation device and the electronic equipment for the abnormal region of the blood vessel, through acquiring the point cloud domain image of the blood vessel image to be segmented, the first segmentation module in the target segmentation model can be utilized to detect and identify each point cloud in the point cloud domain image, whether each point cloud is a blood vessel point or an abnormal point is determined, namely, the segmentation image of the initial abnormal region is obtained, the boundary between the blood vessel region and the initial abnormal region is determined, the point cloud near the boundary is further optimized/blocked by utilizing the second segmentation module, the abnormal point is determined accurately, and the segmentation image of the final abnormal region is obtained accurately.
The foregoing description is only an overview of the present invention, and is intended to be implemented in accordance with the teachings of the present invention in order that the same may be more clearly understood and to make the same and other objects, features and advantages of the present invention more readily apparent.
Drawings
Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention. Also, like reference numerals are used to designate like parts throughout the figures. In the drawings:
FIG. 1 is a flow chart of a method for segmenting an abnormal region of a blood vessel according to an embodiment of the present application;
FIG. 2 is a CTA image;
FIG. 3 is a point cloud domain image obtained by conversion;
FIG. 4 is an enlarged view of a partial point cloud domain image;
FIG. 5 (a) is a sample point cloud image before labeling;
FIG. 5 (b) is a labeled sample point cloud domain image;
FIG. 6 is a schematic diagram illustrating a positional relationship between a boundary region and a region of interest according to an embodiment of the present application;
fig. 7 is a schematic diagram of each blood vessel segment image after image segmentation processing is performed on the point cloud domain image in the embodiment of the present application;
FIG. 8 (a) is an intracranial aneurysm CAT image;
FIG. 8 (b) is a segmented image of an aneurysm obtained by using a conventional manual labeling method for FIG. 8 (a);
fig. 8 (c) is an aneurysm segmentation map obtained by the segmentation method of the abnormal region of the blood vessel in the present application for fig. 8 (a);
fig. 8 (D) is a 3D reconstruction map corresponding to fig. 8 (b);
fig. 8 (e) is a 3D reconstruction map corresponding to fig. 8 (c);
fig. 9 is a block diagram showing a structure of a segmentation apparatus for an abnormal region of a blood vessel according to another embodiment of the present application.
Detailed Description
Various aspects and features of the present application are described herein with reference to the accompanying drawings.
It should be understood that various modifications may be made to the embodiments of the application herein. Therefore, the above description should not be taken as limiting, but merely as exemplification of the embodiments. Other modifications within the scope and spirit of this application will occur to those skilled in the art.
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments of the application and, together with a general description of the application given above and the detailed description of the embodiments given below, serve to explain the principles of the application.
These and other characteristics of the present application will become apparent from the following description of a preferred form of embodiment, given as a non-limiting example, with reference to the accompanying drawings.
It is also to be understood that, although the present application has been described with reference to some specific examples, those skilled in the art can certainly realize many other equivalent forms of the present application.
The foregoing and other aspects, features, and advantages of the present application will become more apparent in light of the following detailed description when taken in conjunction with the accompanying drawings.
Specific embodiments of the present application will be described hereinafter with reference to the accompanying drawings; however, it is to be understood that the disclosed embodiments are merely exemplary of the application, which can be embodied in various forms. Well-known and/or repeated functions and constructions are not described in detail to avoid obscuring the application with unnecessary or excessive detail. Therefore, specific structural and functional details disclosed herein are not intended to be limiting, but merely serve as a basis for the claims and as a representative basis for teaching one skilled in the art to variously employ the present application in virtually any appropriately detailed structure.
The specification may use the word "in one embodiment," "in another embodiment," "in yet another embodiment," or "in other embodiments," which may each refer to one or more of the same or different embodiments as per the application.
The application provides a segmentation method of an abnormal region of a blood vessel, as shown in fig. 1, comprising the following steps:
step S101, acquiring a blood vessel image to be segmented, wherein the blood vessel image comprises a blood vessel region and an abnormal region attached to the blood vessel;
in this step, the abnormal region refers to a focal region of a non-vascular region, and for example, the abnormal region may be an aneurysm region or the like. In this step, the image to be segmented may be any image of a non-point cloud domain type, that is, the image to be segmented may be any mode image, for example, a CT blood vessel imaging (Computed Tomographic Angiography, english: CTA) image as shown in fig. 2, or an MRA image, PET and CT fusion image. Hereinafter, an example of the embodiment of the present application will be described with the abnormal region as an aneurysm.
Step S102, acquiring a point cloud domain image containing the blood vessel region and the abnormal region based on the blood vessel image;
the vessel image to be segmented can be converted after being obtained in this step to obtain a point cloud domain image, for example, a point cloud domain image shown in fig. 3 can be obtained, and a partial enlarged view of the point cloud domain image 3 can be shown in fig. 4, where x in fig. 4 represents an abnormal region, i.e., an aneurysm region.
In the implementation process, the conversion of the image type can be realized by extracting the surface pixel points of the blood vessel region in the CTA image. Of course, other modes of conversion may be adopted, for example, a mode of gridding the image domain data, or a mode of removing non-boundary points in the image domain data, etc. to perform image conversion to obtain the point cloud domain image.
Step S103, inputting the point cloud domain image into a pre-trained target segmentation model to acquire a segmentation image of the abnormal region, wherein the target segmentation model comprises a first segmentation module for segmenting the abnormal region and a second segmentation module for blocking a region adjacent to a boundary between the blood vessel region and the abnormal region;
in the specific implementation process, the first segmentation module in the target segmentation model can be utilized to identify and judge each point cloud/three-dimensional point of the blood vessel image, and each point cloud is determined to be a blood vessel point or an abnormal point, so that an initial abnormal region containing a plurality of abnormal points can be determined based on the identification result of each point cloud. The second segmentation module is further utilized to block point clouds in preset areas on two sides of a boundary between the abnormal area and the blood vessel, and whether the point clouds in the preset areas on two sides of the boundary are blood vessel points or abnormal points is accurately determined, so that the abnormal area can be judged more accurately. Wherein the segmented image may be a point cloud domain image.
According to the segmentation method of the abnormal region of the blood vessel, through obtaining the point cloud domain image of the blood vessel image to be segmented, the first segmentation module in the target segmentation model can be utilized to detect and identify each point cloud in the point cloud domain image, whether each point cloud is a blood vessel point or an abnormal point is determined, namely, the segmentation image of the initial abnormal region is obtained, the boundary between the blood vessel region and the initial abnormal region is determined, further optimization/blocking of the point cloud near the boundary is facilitated by utilizing the second segmentation module, and therefore the abnormal point is accurately determined, and further the segmentation image of the final abnormal region is accurately obtained.
Taking a vascular aneurysm as an example, the related art has less supervision on the boundary due to the small area occupation ratio, so that the effect of boundary segmentation is not good, and for medical image segmentation, particularly for aneurysm segmentation tasks, the boundary segmentation is of great importance, which greatly guides the position of aneurysm cutting. The target segmentation model of the embodiment of the application not only can realize segmentation of the blood vessel and the aneurysm, but also can optimize the junction between the blood vessel and the aneurysm, thereby accurately determining the aneurysm.
In another embodiment of the present application, a method for segmenting an abnormal region of a blood vessel is provided, and on the basis of the foregoing embodiment, in this embodiment, when obtaining a point cloud domain image based on a transformation of a blood vessel image, specifically any one of the following modes may be adopted:
mode one: carrying out gridding treatment on blood vessels and abnormal areas in the blood vessel image, and extracting at least one pixel point in the grid to obtain a pixel point set; and converting the pixel point set into a point cloud domain, and acquiring a point cloud domain image.
Mode two: extracting boundary point pixels of a blood vessel region and an abnormal region in the blood vessel image to obtain a pixel point set; and converting the pixel point set into a point cloud domain, and acquiring a point cloud domain image.
In this embodiment, the blood vessel image of the point cloud domain is obtained by processing the blood vessel image of the non-point cloud domain, so that the anatomical feature of the abnormal region can be fully utilized, the convex shape of the abnormal region can be clearly shown in the point cloud domain image, and the foundation is laid for the follow-up accurate determination of the range of the abnormal region based on the convex feature of the point cloud domain, thereby accurately dividing the image of the abnormal region from the point cloud domain image. The conversion of the pixel point set into the point cloud domain is a coordinate conversion relationship, that is, the pixel point set is converted from an image coordinate system into a point cloud coordinate system, which may be a world coordinate system.
In this embodiment, when the point cloud domain image is input into a pre-trained target segmentation model, and a segmented image of the abnormal region is obtained, the specific process is as follows: acquiring an initial segmentation image of the abnormal region based on the first segmentation module; and performing barrier processing on point clouds in specific areas on two sides of a boundary between the blood vessel and the abnormal area based on the second segmentation module to obtain segmented images of the abnormal area, wherein the barrier processing comprises classifying the point clouds on one side of the abnormal area as the abnormal area and classifying the point clouds on one side of the blood vessel area as the blood vessel area.
That is, feature extraction may be performed on each point cloud/three-dimensional point of the blood vessel image by using the first segmentation module in the target segmentation model to obtain feature information of each point cloud, and then each point cloud is determined to be a blood vessel point or an abnormal point based on the feature information of each point cloud, so that an initial abnormal region including a plurality of abnormal points may be determined based on a determination result of each point cloud, and a boundary between the initial abnormal region and the blood vessel region may be determined, and it is noted that an output result of the first segmentation module at a position adjacent to the boundary may be inaccurate due to insufficient attention to the point cloud at the boundary position. And then, the point clouds in the preset areas at the two sides of the boundary can be further optimized by using the second segmentation module, so that whether the point clouds in the preset areas at the two sides of the boundary are vascular points or abnormal points can be secondarily judged, the judgment of the abnormal areas can be more accurate, and a guarantee is provided for the follow-up accurate segmentation of the abnormal areas from the point cloud domain images.
In order to reduce the operation amount during the segmentation of the abnormal region, the segmentation processing may be further performed on the point cloud domain image before the point cloud domain image is input into a pre-trained target segmentation model to obtain the segmented image of the abnormal region, so as to obtain a plurality of blood vessel segment images, i.e. the point cloud domain image is segmented into a plurality of blood vessel segment images. When the abnormal region is segmented, each blood vessel segment image can be input into a pre-trained target segmentation model, and a plurality of sub-segmentation images of the abnormal region corresponding to a plurality of blood vessel segments are obtained; and finally, recombining the sub-divided images to obtain the divided images of the abnormal region.
That is, for each blood vessel segment image, the first segmentation module in the target segmentation model is utilized to perform feature extraction on each point cloud/three-dimensional point of the blood vessel segment image, so as to obtain feature information of each point cloud, and then determine that each point cloud is a blood vessel point or an abnormal point based on the feature information of each point cloud, so that an initial abnormal region containing a plurality of abnormal points can be determined based on a judgment result of each point cloud, and a boundary between the initial abnormal region and the blood vessel region is determined. And then, further optimizing the point clouds in the preset areas at the two sides of the boundary by using the second segmentation module, so as to secondarily judge whether the point clouds in the preset areas at the two sides of the boundary are vascular points or abnormal points. Thus, an abnormal region including each abnormal point can be obtained according to the final recognition result of the point cloud. Of course, when the identification results of each point cloud in the same blood vessel segment image are all blood vessel points, determining that the blood vessel segment image does not contain an abnormal region; when the identification results of each point cloud in the same blood vessel segment image are all abnormal points or part of the point cloud is an abnormal point, determining that the blood vessel segment contains an abnormal movement region.
In this embodiment, after obtaining the sub-divided images of the abnormal region corresponding to each vessel segment image, it may be determined that the vessel segment image including the abnormal region is the target vessel segment image, then determine the first position information of the sub-divided image in the target vessel segment image and the second position information of the target vessel segment in the point cloud domain image, and determine the target position information of the sub-divided image in the point cloud domain image based on the first position information and the second position information, so as to reconstruct each sub-divided image according to the target position information of each divided image, thereby obtaining the final abnormal region divided image. In this embodiment, since there may be multiple abnormal regions corresponding to the same point cloud domain image, by determining that each sub-divided image is located at a target position in the point cloud domain image, the recombination of the sub-divided images can be more accurate, and a guarantee is provided for accurately obtaining a final divided image of the abnormal region.
Still another embodiment of the present application provides a segmentation method of an abnormal region of a blood vessel, including:
step S201, training to obtain a target segmentation model;
in the specific implementation process, the training process of the target segmentation model is as follows:
Step 2011, acquiring a sample point cloud domain image containing an abnormal region and a blood vessel region and a label point cloud domain image marking the abnormal region;
in the specific implementation process, a plurality of sample images can be acquired; and then converting each acquired sample image, extracting the vascular surface pixel points corresponding to the vascular objects in each sample image, and constructing and acquiring a sample point cloud domain image corresponding to each sample image based on each vascular surface pixel point. Specifically, performing gridding treatment on blood vessels and abnormal areas in the blood vessel image, and extracting at least one pixel point in the grid to obtain a pixel point set; and converting the pixel point set into a point cloud domain, and acquiring a point cloud domain image. Or extracting boundary point pixels of a blood vessel region and an abnormal region in the blood vessel image to obtain a pixel point set; and converting the pixel point set into a point cloud domain, and acquiring a point cloud domain image.
In this step, when obtaining the tag point cloud domain image for labeling the abnormal area, the method specifically may include: acquiring the boundary between a blood vessel and an abnormal region in the sample point cloud domain image and a seed point in the abnormal region; and labeling the abnormal region in the sample point cloud domain image based on the anatomical morphological characteristics, the boundary and the seed points of the abnormal region, and obtaining a label point cloud domain image.
Specifically, the abnormal region labeling can be performed on the point cloud domain image manually. Namely, responding to the labeling operation of a user on each sample point cloud domain image, and acquiring the boundary between a blood vessel and an abnormal region in the sample point cloud domain image and a seed point in the abnormal region; and labeling the abnormal region in the sample point cloud domain image based on the anatomical morphological characteristics, the boundary and the seed points of the abnormal region to obtain a label point cloud domain image. The labeling process is performed on the point cloud data, and the labeling of the abnormal region on the point cloud domain image has three advantages: first, on the point cloud domain image, the expression form of the abnormal region is clearer and easier to identify. Secondly, considering that the abnormal region is of a raised closed loop structure, only the junction of the abnormal region and the blood vessel and a seed point of the abnormal region can be marked by using an image processing algorithm, and all marking results are automatically generated by the image processing algorithm, so that the marking workload is greatly reduced. Thirdly, by adopting the marking method, not only the abnormal region but also the connection junction region of the abnormal region and the blood vessel can be marked, so that finer optimization of the junction is possible. For example, for a sample point cloud domain image before labeling as shown in fig. 5 (a), the sample point cloud domain image after labeling can be shown in fig. 5 (b), the user only needs to label a connecting line y in the graph and a seed point x of one aneurysm, and all the aneurysm areas can be labeled by using an image processing algorithm. By adopting the labeling method, the labeling speed is faster, the labeling area is more accurate, the available labeling types are more, and the labeling of the aneurysm area and the label at the junction of the aneurysm and the blood vessel can be obtained simultaneously.
Step 2012, training an initial first segmentation module based on the sample point cloud domain image and the label point cloud domain image, and acquiring the first segmentation module when a first preset termination condition is reached;
in the implementation process, the first segmentation module in the initial segmentation model can segment the abnormal region of each sample point cloud domain image to obtain a segmentation result of each sample point cloud domain image; and adjusting model parameters in the first segmentation module based on the difference between the segmentation result of each sample point cloud domain image and the label information until a first preset termination condition is met, and stopping training to obtain the first segmentation module. The first preset termination adjustment includes a value of a loss function corresponding to the first segmentation module reaching a preset threshold or a number of training of the first segmentation module reaching a preset number.
The method comprises the following steps: labeling can be performed on sample point cloud domain data which is not segmented; after labeling, segmenting, taking each blood vessel segment as a training sample, namely taking a blood vessel segment image containing an abnormal region as a positive sample and taking a blood vessel segment image not containing the abnormal region as a negative sample, wherein the representation form can be coordinate values, but is not limited to the coordinate values, and can also be the characteristics of normal vector and the like, inputting the positive sample and the negative sample into the initial first segmentation module, carrying out characteristic extraction on each three-dimensional point/point cloud of each sample blood vessel segment image by utilizing the initial first segmentation module to obtain characteristic information of each sample point, and carrying out abnormal point identification on the characteristic information of each sample point to obtain a first identification result of each sample point so as to obtain an abnormal region segmentation result of each sample point cloud domain image; and finally, determining differences between the segmentation results of the sample point cloud domain images and the label information corresponding to the sample point cloud domain images based on the segmentation results of the abnormal areas of the sample point cloud domain images and the label information of the abnormal areas contained in the sample point cloud domain images, and adjusting parameters in the initial first segmentation module based on the differences until training conditions are met, so as to obtain the first segmentation module. The network structure of the first segmentation module in this step may employ an automatic coding network in a form similar to the uiet (uiet is a segmentation network applied to medical influence proposed on the basis of Fcn), but is not limited thereto. The final first segmentation module can be obtained by optimizing the loss function.
Step S2013, acquiring the boundary between the abnormal region and the blood vessel region based on the label point cloud domain image;
for example, as shown in fig. 6, first, the boundary region Ω between the abnormal region and the blood vessel region may be determined according to the labeling result; the area on the interface side can then be denoted as Ω + The other side area is denoted by Ω - 。
Step S2014, training an initial second segmentation module based on the boundary and tag point cloud domain image and the output image of the first segmentation module, and acquiring a second segmentation module when a second preset termination condition is reached, wherein a region of interest of the second segmentation module at least comprises regions in specific ranges on two sides of the boundary;
in this step, the boundary region between the abnormal region and the blood vessel region in the blood vessel segment image/point cloud domain image is determined for the positive sample including the abnormal region. Regions of interest on either side of the interface region are then determined based on the interface region. Acquiring characteristic information of each sample point cloud in the attention area; in this step, the feature information of each sample point cloud may be specifically obtained by extracting through the first segmentation submodule. The first segmentation module extracts the obtained characteristic information of each sample point cloud in the concerned region, inputs the characteristic information into the initial second segmentation module, then further utilizes the initial second segmentation module to carry out secondary identification on each sample point cloud in the concerned region to obtain a second identification result of each sample point cloud, so as to obtain a current abnormal point identification result corresponding to the point cloud image/the vessel segment image based on the second identification result of each point cloud in the concerned region in the same point cloud image/the vessel segment image and the identification result of each point cloud in the non-concerned region, and finally adjusts parameters in the initial second segmentation module based on the difference between the identification result of the current abnormal point and label information corresponding to each point cloud image/the vessel segment image to obtain a final second segmentation module. The second preset termination adjustment includes the value of the loss function corresponding to the second segmentation module reaching a preset threshold or the number of training of the second segmentation module reaching a preset number.
In a specific implementation, the region of interest may be determined by a nearest neighbor algorithm and a fixed threshold (width to distance intersection), i.e., Ω in fig. 6, respectively + And omega - In the region, the iterative determination of the region of interest A relatively close to the boundary + And region of interest A - . Of course, the region of interest A + And region of interest A - Or may be manually selected. Finally, the sample point clouds in the two parts of the concerned areas can be detected for the second time, so that samples which are easy to predict errors in the parts are optimized, and the second segmentation module is obtained through training. The specific training process is as follows: and taking the characteristic information learned by the first segmentation module as input. The second segmentation module starts from the region of interest A + The prediction results/recognition results of the sample points should be kept consistent, and the region A of interest is the same - The prediction result/recognition result of the sample points in the map should be consistent, and the region of interest A should be kept + With the region of interest A - The results of each sample point cloud within should be opposite. That is, due to the existence of the boundary region, A is blocked + And A - The connection of the sample point clouds in the middle, that is, the samples in the perception domain are necessarily the same class samples in the training process of the model, so that the segmentation effect can be greatly enhanced, and the parameters of the initial second segmentation module are adjusted based on the difference between the identification obtained result of the initial second segmentation module and the label information of each sample point cloud in the concerned region until the training strip is satisfied And stopping training when the workpiece is finished, and obtaining a final second segmentation module.
In this step, after the first segmentation module and the second segmentation module are obtained, the target segmentation model can be constructed and obtained based on the two modules.
Step S202, acquiring a point cloud domain image containing the blood vessel region and the abnormal region;
in the step, the blood vessel and the abnormal region in the blood vessel image can be subjected to gridding treatment, and at least one pixel point in the grid is extracted to obtain a pixel point set; and converting the pixel point set into a point cloud domain, and acquiring a point cloud domain image. Or extracting boundary point pixels of a blood vessel region and an abnormal region in the blood vessel image to obtain a pixel point set; and converting the pixel point set into a point cloud domain, and acquiring a point cloud domain image. By extracting the surface pixel points of the blood vessel object and constructing a point cloud domain image based on the surface pixel points, the point cloud domain image containing the surface characteristics of the blood vessel can be sufficiently obtained, and a foundation is laid for the subsequent segmentation of abnormal regions of the blood vessel based on the point cloud domain image.
Step S203, dividing the point cloud domain image into a plurality of blood vessel segment images;
in this step, the following manner may be specifically adopted when the segmentation processing/segmentation processing is performed: n points are uniformly sampled on the whole point cloud image. And then each point is taken as an initial point, all points with the distance from the ground to the initial point being smaller than a certain threshold value are recorded, and then the points and the initial point together form a blood vessel segment. Similarly, an overall point cloud image may be divided into N vessel segments.
In this step, after image segmentation is performed, each blood vessel segment image obtained by segmentation may be as shown in fig. 7.
Step S204, inputting each blood vessel segment image into a pre-trained target segmentation model, and obtaining a plurality of sub-segmentation images of an abnormal region corresponding to a plurality of blood vessel segments; recombining a plurality of sub-divided images to obtain a divided image of the abnormal region;
in this step, feature extraction is performed on each point cloud/three-dimensional point of the blood vessel segment image by using a first segmentation module in the target segmentation model for each blood vessel segment image, so as to obtain feature information of each point cloud, and then each point cloud is determined to be a blood vessel point or an abnormal point based on the feature information of each point cloud, so that an initial abnormal region including a plurality of abnormal points can be determined based on a first judgment result/a first identification result of each point cloud, and a boundary between the initial abnormal region and the blood vessel region can be determined. And then, further optimizing the point clouds in the preset areas at the two sides of the boundary by using a second segmentation module, so as to secondarily judge whether the point clouds in the preset areas (the concerned areas) at the two sides of the boundary are vascular points or abnormal points, and obtaining a second recognition result of the point clouds in the concerned areas. Thus, an abnormal region including each abnormal point can be obtained according to the final recognition result of the point cloud. Of course, when the identification results of each point cloud in the same blood vessel segment image are all blood vessel points, determining that the blood vessel segment image does not contain an abnormal region; when the identification results of each point cloud in the same blood vessel segment image are all abnormal points or part of the point cloud is an abnormal point, determining that the blood vessel segment contains an abnormal movement region.
In this embodiment, after obtaining the sub-divided images of the abnormal region corresponding to each vessel segment image, it may be determined that the vessel segment image including the abnormal region is the target vessel segment image, then determine the first position information of the sub-divided image in the target vessel segment image and the second position information of the target vessel segment in the point cloud domain image, and determine the target position information of the sub-divided image in the point cloud domain image based on the first position information and the second position information, so as to reconstruct each sub-divided image according to the target position information of each divided image, thereby obtaining the final abnormal region divided image. In this embodiment, since there may be multiple abnormal regions corresponding to the same point cloud domain image, by determining that each sub-divided image is located at a target position in the point cloud domain image, the recombination of the sub-divided images can be more accurate, and a guarantee is provided for accurately obtaining a final divided image of the abnormal region.
Step S205, displaying the abnormal region and the blood vessel region in the point cloud image by using different display modes.
In this step, after the sub-divided images are recombined, the position of the abnormal region in the point cloud domain image may be further determined, and then the abnormal region and the blood vessel region are displayed in different display modes based on the position. Alternatively, the abnormal region and the blood vessel region in each blood vessel segment image may be displayed in different display modes.
In the implementation process, the method can be directly displayed based on the result of the point cloud domain, and the display method can be directly output to an application needing 3D reconstruction of an image, such as modeling, operation and the like of a meta universe. In the step, the abnormal region is located at the position in the point cloud domain image and displayed in different display modes, so that a user can know the specific position of the aneurysm more clearly, and help is provided for subsequent diagnosis and treatment.
In addition, in this embodiment, after determining the abnormal area or the blood vessel area, the result of the point cloud domain may be further post-processed to restore to the image domain, so as to obtain the result of the non-point cloud domain image. That is, the object segmentation model further comprises a post-processing module for restoring the output result of the second segmentation module to the image domain, i.e. segmenting the image into image domain images. And determining a first region corresponding to the abnormal region and a second region corresponding to the blood vessel region in the non-point cloud domain image based on the position coordinates of the abnormal region in the point cloud domain image, and finally displaying the first region and the second region based on different display modes.
The abnormal region detection method in the embodiment has higher sensitivity and lower false positive: this is because the present application makes full use of the anatomical feature of the abnormal region, and can clearly show the shape change of the protrusion thereof in the point cloud domain, so that the segmentation accuracy is higher.
The abnormal region segmentation method in the application is more generalized: the method in the application does not need an original CTA image, only needs a segmentation result diagram of a blood vessel, so that the segmentation effect of the aneurysm is not influenced by image contrast, and the influence of imaging differences of different modes and equipment of different manufacturers on the aneurysm is small.
The aneurysm segmentation method in the application requires fewer computational resources: the point cloud domain data only collects points on the surface of the region of interest, and the points are subjected to downsampling in the training and application stage, so that the number of sample points actually participating in calculation is less, and the consumption of calculation resources is less.
The aneurysm segmentation method in the application can meet clinical requirements: because of the function of the abnormal region juncture optimization module, the segmentation accuracy of the abnormal region segmentation result obtained by the method is higher at the position connected with the blood vessel, which is the most focused position of clinical tasks such as aneurysm excision operation.
In the present application, for the intracranial aneurysm CAT image shown in fig. 8 (a), by performing aneurysm detection by using the abnormal region segmentation method in the present application, an aneurysm segmentation map and a corresponding 3D reconstruction map as shown in fig. (c) and (e) can be obtained. Wherein, the diagrams (b) and (D) are aneurysm segmentation diagrams and corresponding 3D reconstruction diagrams obtained by adopting a traditional manual marking method. The comparison shows that the obtained segmentation result is accurate in aneurysm identification and clear in boundary segmentation, so that the method can meet the clinical application requirements and has good practicability.
Another embodiment of the present application provides a segmentation apparatus for an abnormal region of a blood vessel, as shown in fig. 9, including:
a first acquisition module 11 for acquiring a blood vessel image to be segmented, the blood vessel image including a blood vessel region and an abnormal region attached to the blood vessel;
a second acquisition module 12, configured to acquire a point cloud domain image including the blood vessel region and the abnormal region based on the blood vessel image;
a segmentation module 13, configured to input the point cloud domain image into a pre-trained target segmentation model, and acquire a segmentation image of the abnormal region, where the target segmentation model includes a first segmentation module for segmenting the abnormal region and a second segmentation module for blocking a region adjacent to a boundary between the blood vessel region and the abnormal region.
In a specific implementation process of this embodiment, the first obtaining module is specifically configured to: carrying out gridding treatment on blood vessels and abnormal areas in the blood vessel image, and extracting at least one pixel point in the grid to obtain a pixel point set; converting the pixel point set into a point cloud domain, and acquiring a point cloud domain image; or extracting boundary point pixels of a blood vessel region and an abnormal region in the blood vessel image to obtain a pixel point set; and converting the pixel point set into a point cloud domain, and acquiring a point cloud domain image.
In a specific implementation process of this embodiment, the segmentation module is specifically configured to: acquiring an initial segmentation image of the abnormal region based on the first segmentation module; and performing barrier treatment on point clouds in specific areas on two sides of a boundary between a blood vessel and an abnormal area in the initial segmented image based on the second segmentation module to obtain segmented images of the abnormal area, wherein the barrier treatment comprises classifying the point clouds on one side of the abnormal area as the abnormal area and classifying the point clouds on one side of the blood vessel area as the blood vessel area.
In a specific implementation process of this embodiment, the segmentation apparatus for an abnormal region of a blood vessel further includes a segmentation module, where the segmentation module is configured to: before inputting the point cloud domain image into a pre-trained target segmentation model and acquiring a segmentation image of the abnormal region, segmenting the point cloud domain image into a plurality of blood vessel segment images;
the segmentation module is specifically used for: inputting the plurality of blood vessel segment images into a pre-trained target segmentation model to obtain a plurality of sub-segmentation images of an abnormal region corresponding to the plurality of blood vessel segments; and recombining the sub-divided images to obtain the divided images of the abnormal region.
In a specific implementation process of this embodiment, the segmentation apparatus for an abnormal region of a blood vessel further includes a training module, where the training module is configured to: acquiring a sample point cloud domain image containing an abnormal region and a blood vessel and a label point cloud domain image marking the abnormal region; training an initial first segmentation module based on the sample point cloud domain image and the label point cloud domain image, and acquiring the first segmentation module when a first preset termination condition is reached; acquiring the boundary between the abnormal region and the blood vessel based on the label point cloud domain image; and training an initial second segmentation module based on the boundary and tag point cloud domain image and the output image of the first segmentation module, and acquiring the second segmentation module when a second preset termination condition is reached, wherein the concerned region of the second segmentation module at least comprises regions in specific ranges on two sides of the boundary.
In a specific implementation process of this embodiment, the training module is specifically configured to: and determining the region of interest of the second segmentation module through a neighbor algorithm or a fixed threshold method.
In a specific implementation process of this embodiment, the training module is specifically configured to: acquiring the boundary between a blood vessel and an abnormal region in the sample point cloud domain image and a seed point in the abnormal region;
And labeling the abnormal region in the sample point cloud domain image based on the anatomical morphological characteristics, the boundary and the seed points of the abnormal region, and obtaining a label point cloud domain image.
In a specific implementation process of this embodiment, the segmentation apparatus for an abnormal region of a blood vessel further includes a display module, where the display module is configured to: and displaying the abnormal region and the blood vessel region in the point cloud domain image by adopting different display modes.
According to the segmentation device for the abnormal region of the blood vessel, the point cloud domain image of the blood vessel image to be segmented is obtained, the first segmentation module in the target segmentation model can be utilized to detect and identify each point cloud in the point cloud domain image, whether each point cloud is a blood vessel point or an abnormal point is determined, the segmentation image of the initial abnormal region is obtained, the juncture of the blood vessel region and the initial abnormal region is determined, the point cloud near the juncture is further optimized/blocked by the second segmentation module conveniently, the abnormal point is determined accurately, and the segmentation image of the final abnormal region is obtained accurately.
Another embodiment of the present application provides an electronic device, at least including a memory, and a processor, where the memory stores a computer program, and the processor when executing the computer program on the memory implements the following method steps:
Step one, acquiring a blood vessel image to be segmented, wherein the blood vessel image comprises a blood vessel region and an abnormal region attached to the blood vessel;
step two, acquiring a point cloud domain image comprising the blood vessel region and the abnormal region based on the blood vessel image;
inputting the point cloud domain image into a pre-trained target segmentation model to acquire a segmentation image of the abnormal region, wherein the target segmentation model comprises a first segmentation module for segmenting the abnormal region and a second segmentation module for blocking a region adjacent to the boundary between the blood vessel region and the abnormal region.
The specific implementation process of the above method steps can refer to the embodiment of the above method for segmenting an abnormal region of any blood vessel, and this embodiment is not repeated here.
According to the electronic device in the embodiment, through acquiring the point cloud domain image of the blood vessel image to be segmented, the first segmentation module in the target segmentation model can be utilized to detect and identify each point cloud in the point cloud domain image, whether each point cloud is a blood vessel point or an abnormal point is determined, namely, the segmented image of the initial abnormal region is acquired, the juncture of the blood vessel region and the initial abnormal region is determined, further optimization/blocking of the point cloud near the juncture is facilitated by utilizing the second segmentation module, the abnormal point is determined accurately, and the segmented image of the final abnormal region is acquired accurately.
The above embodiments are only exemplary embodiments of the present application and are not intended to limit the present application, the scope of which is defined by the claims. Various modifications and equivalent arrangements may be made to the present application by those skilled in the art, which modifications and equivalents are also considered to be within the scope of the present application.
Claims (10)
1. A method for segmenting an abnormal region of a blood vessel, comprising:
acquiring a blood vessel image to be segmented, wherein the blood vessel image comprises a blood vessel region and an abnormal region attached to the blood vessel;
acquiring a point cloud domain image containing the blood vessel region and the abnormal region based on the blood vessel image;
inputting the point cloud domain image into a pre-trained target segmentation model to acquire a segmentation image of the abnormal region, wherein the target segmentation model comprises a first segmentation module for segmenting the abnormal region and a second segmentation module for blocking a region adjacent to a boundary between the blood vessel region and the abnormal region.
2. The method of claim 1, wherein the acquiring a point cloud domain image containing the vessel region and the anomaly region based on the vessel image comprises:
Carrying out gridding treatment on blood vessels and abnormal areas in the blood vessel image, and extracting at least one pixel point in the grid to obtain a pixel point set;
converting the pixel point set into a point cloud domain, and acquiring a point cloud domain image; or alternatively, the first and second heat exchangers may be,
extracting boundary point pixels of a blood vessel region and an abnormal region in the blood vessel image to obtain a pixel point set;
and converting the pixel point set into a point cloud domain, and acquiring a point cloud domain image.
3. The method of claim 1, wherein the inputting the point cloud domain image into a pre-trained target segmentation model, obtaining a segmented image of the anomaly region, comprises:
acquiring an initial segmentation image of the abnormal region based on the first segmentation module;
and performing blocking processing on point clouds in specific areas on two sides of a boundary between an abnormal area and a blood vessel area in the initial segmented image based on the second segmentation module, and acquiring segmented images of the abnormal area, wherein the blocking processing comprises classifying the point clouds on one side of the abnormal area as the abnormal area and classifying the point clouds on one side of the blood vessel area as the blood vessel area.
4. The method of claim 1, further comprising, prior to inputting the point cloud domain image into a pre-trained target segmentation model, acquiring a segmented image of the anomaly region:
Dividing the point cloud domain image into a plurality of blood vessel segment images;
inputting the point cloud domain image into a pre-trained target segmentation model to obtain a segmented image of the abnormal region, wherein the method comprises the following steps:
inputting the plurality of blood vessel segment images into a pre-trained target segmentation model to obtain a plurality of sub-segmentation images of an abnormal region corresponding to the plurality of blood vessel segments;
and recombining the sub-divided images to obtain the divided image of the abnormal region.
5. The method as recited in claim 1, further comprising: obtaining a pre-trained target segmentation model, wherein the obtaining the pre-trained target segmentation model comprises:
acquiring a sample point cloud domain image containing an abnormal region and a blood vessel and a label point cloud domain image marking the abnormal region;
training an initial first segmentation module based on the sample point cloud domain image and the label point cloud domain image, and acquiring the first segmentation module when a first preset termination condition is reached;
acquiring the boundary between the abnormal region and the blood vessel based on the label point cloud domain image;
and training an initial second segmentation module based on the boundary and tag point cloud domain image and the output image of the first segmentation module, and acquiring the second segmentation module when a second preset termination condition is reached, wherein the concerned region of the second segmentation module at least comprises regions in specific ranges on two sides of the boundary, the first preset termination condition comprises the condition that the value of a loss function corresponding to the first segmentation module reaches a preset threshold value or the training times of the first segmentation module reaches preset times, and the second preset termination condition comprises the condition that the value of the loss function corresponding to the second segmentation module reaches the preset threshold value or the training times of the second segmentation module reaches preset times.
6. The method as recited in claim 5, further comprising: and determining the region of interest of the second segmentation module through a neighbor algorithm or a fixed threshold method.
7. The method of claim 5, wherein the obtaining a tag point cloud domain image that labels the anomaly region comprises:
acquiring the boundary between a blood vessel and an abnormal region in the sample point cloud domain image and a seed point in the abnormal region;
and labeling the abnormal region in the sample point cloud domain image based on the anatomical morphological characteristics, the boundary and the seed points of the abnormal region, and obtaining a label point cloud domain image.
8. The method of claim 1, wherein the object segmentation model further comprises a post-processing module for restoring the output of the second segmentation module to the image domain.
9. A segmentation apparatus for an abnormal region of a blood vessel, comprising:
the first acquisition module is used for acquiring a blood vessel image to be segmented, wherein the blood vessel image comprises a blood vessel region and an abnormal region attached to the blood vessel;
the second acquisition module is used for acquiring a point cloud domain image containing the blood vessel region and the abnormal region based on the blood vessel image;
The segmentation module is used for inputting the point cloud domain image into a pre-trained target segmentation model to acquire a segmentation image of the abnormal region, wherein the target segmentation model comprises a first segmentation module used for segmenting the abnormal region and a second segmentation module used for blocking a region adjacent to the boundary between the blood vessel region and the abnormal region.
10. An electronic device comprising at least a memory, a processor, the memory having stored thereon a computer program, the processor, when executing the computer program on the memory, performing the steps of the method for segmenting an abnormal region of a blood vessel according to any one of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310347024.3A CN116563305A (en) | 2023-04-03 | 2023-04-03 | Segmentation method and device for abnormal region of blood vessel and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310347024.3A CN116563305A (en) | 2023-04-03 | 2023-04-03 | Segmentation method and device for abnormal region of blood vessel and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116563305A true CN116563305A (en) | 2023-08-08 |
Family
ID=87485173
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310347024.3A Pending CN116563305A (en) | 2023-04-03 | 2023-04-03 | Segmentation method and device for abnormal region of blood vessel and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116563305A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117058464A (en) * | 2023-08-31 | 2023-11-14 | 强联智创（北京）科技有限公司 | Method and device for training generation model for generating healthy blood vessel surface |
-
2023
- 2023-04-03 CN CN202310347024.3A patent/CN116563305A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117058464A (en) * | 2023-08-31 | 2023-11-14 | 强联智创（北京）科技有限公司 | Method and device for training generation model for generating healthy blood vessel surface |
| CN117058464B (en) * | 2023-08-31 | 2024-06-11 | 强联智创（北京）科技有限公司 | Method and device for training generation model for generating healthy blood vessel surface |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110111313B (en) | Medical image detection method based on deep learning and related equipment | |
| CN107784647B (en) | Liver and tumor segmentation method and system based on multitask deep convolutional network | |
| CN110232383B (en) | Focus image recognition method and focus image recognition system based on deep learning model | |
| CN110428432B (en) | Deep neural network algorithm for automatically segmenting colon gland image | |
| CN111445478B (en) | Automatic intracranial aneurysm region detection system and detection method for CTA image | |
| CN109522908B (en) | Image significance detection method based on region label fusion | |
| US20220198230A1 (en) | Auxiliary detection method and image recognition method for rib fractures based on deep learning | |
| Liu et al. | A framework of wound segmentation based on deep convolutional networks | |
| CN110675407B (en) | Image instance segmentation method and device, electronic equipment and storage medium | |
| CN108830149B (en) | Target bacterium detection method and terminal equipment | |
| CN110974306B (en) | System for discernment and location pancreas neuroendocrine tumour under ultrasonic endoscope | |
| CN112308846B (en) | Blood vessel segmentation method and device and electronic equipment | |
| CN112446892A (en) | Cell nucleus segmentation method based on attention learning | |
| CN114937086B (en) | Training method and detection method for multi-image target detection and related products | |
| CN112991365B (en) | Coronary artery segmentation method, system and storage medium | |
| CN112037146A (en) | Medical image artifact automatic correction method and device and computer equipment | |
| Wang et al. | Left atrial appendage segmentation based on ranking 2-D segmentation proposals | |
| CN111932495B (en) | Medical image detection method, device and storage medium | |
| CN111415364A (en) | Method, system and storage medium for converting image segmentation samples in computer vision | |
| CN116563305A (en) | Segmentation method and device for abnormal region of blood vessel and electronic equipment | |
| CN116883341A (en) | Liver tumor CT image automatic segmentation method based on deep learning | |
| CN112991280B (en) | Visual detection method, visual detection system and electronic equipment | |
| CN111401102A (en) | Deep learning model training method and device, electronic equipment and storage medium | |
| CN116958679A (en) | Target detection method based on weak supervision and related equipment | |
| CN116342446A (en) | Multi-focus image fusion method and device, electronic equipment 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 |