CN111325759A

CN111325759A - Blood vessel segmentation method, device, computer equipment and readable storage medium

Info

Publication number: CN111325759A
Application number: CN202010177262.0A
Authority: CN
Inventors: 宋燕丽; 杨帆; 吴迪嘉
Original assignee: Shanghai United Imaging Intelligent Healthcare Co Ltd
Current assignee: Lianying Intelligent Medical Technology Beijing Co ltd; Shanghai United Imaging Intelligent Healthcare Co Ltd
Priority date: 2020-03-13
Filing date: 2020-03-13
Publication date: 2020-06-23
Anticipated expiration: 2040-03-13
Also published as: CN111325759B

Abstract

The application relates to a blood vessel segmentation method, a blood vessel segmentation device, a computer device and a storage medium. The method comprises the following steps: acquiring an image to be detected; inputting the image to be detected into a tissue and organ segmentation network to obtain a tissue and organ image; obtaining a blood vessel and a distance field of the tissue organ corresponding to the blood vessel according to the tissue organ image; determining an annular band containing the blood vessel from the distance field and the tissue organ image; inputting the annular band into the blood vessel segmentation network to obtain an initial blood vessel image; extracting a first image block within a preset range around the center line end point of the blood vessel according to the initial blood vessel image; and inputting the first image block into a blood vessel tracking network for end point tracking to obtain a final blood vessel image. By the method, the accuracy and the stability of the acquired blood vessel image can be improved, the image is corrected through the neural network, the time and the energy of a technician are greatly saved, and the diagnosis cost is further saved.

Description

Blood vessel segmentation method, device, computer equipment and readable storage medium

Technical Field

The present application relates to the field of medical imaging, and in particular, to a blood vessel segmentation method, an apparatus, a computer device, and a readable storage medium.

Background

Cardiovascular diseases are diseases with high morbidity and mortality, wherein common diseases such as coronary artery stenosis and plaques seriously harm human health. Cardiovascular diseases have the characteristics of acute onset, strong concealment and the like, so that the realization of the diagnosis of heart diseases has very important clinical significance. Because coronary vessels are slender tubular structures and have high variability, clinical diagnosis of coronary vessels generally depends on the segmentation result of the coronary vessels, namely, the coronary arteries are automatically segmented, and then the coronary vessels are diagnosed by adopting visualization technologies such as CPR reconstruction (curved surface reconstruction), coronary probe reconstruction and VR (virtual reality) on the basis of segmentation, so that a basis is provided for early prevention and diagnosis of cardiovascular diseases for doctors.

The coronary segmentation techniques commonly used at present are generally based on traditional image processing methods, such as linear structure detection, deformation models, or based on traditional machine learning methods. However, these methods generally have poor accuracy and stability of segmentation, and require a technician to spend a lot of time and effort to perform manual correction and editing, which greatly affects the diagnosis of coronary artery disease by a doctor.

Disclosure of Invention

The embodiment of the application provides a blood vessel segmentation method, a blood vessel segmentation device, computer equipment and a readable storage medium, and aims to at least solve the problem that the accuracy and the stability of related technologies are poor.

A vessel segmentation method comprising: acquiring an image to be detected; inputting the image to be detected into a tissue and organ segmentation network to obtain a tissue and organ image; obtaining a blood vessel and a distance field of the tissue organ corresponding to the blood vessel according to the tissue organ image; determining an annular band containing the blood vessel from the distance field and the tissue organ image; inputting the annular band into the blood vessel segmentation network to obtain an initial blood vessel image; extracting a first image block within a preset range around the center line end point of the blood vessel according to the initial blood vessel image; and inputting the first image block into a blood vessel tracking network for end point tracking to obtain a final blood vessel image.

In one embodiment, inputting the annular band into the vessel segmentation network, and obtaining an initial vessel image includes: dividing the annular band into a plurality of second image blocks according to a preset size; respectively inputting the second image blocks into the blood vessel segmentation network to obtain a plurality of blood vessel segmentation images; and splicing the plurality of blood vessel segmentation images to obtain an initial blood vessel image.

In one embodiment, the stitching the plurality of blood vessel segmentation images to obtain an initial blood vessel image includes: the second image block comprises a position code; acquiring the position code of each blood vessel segmentation image in an annular band according to the position code of each second image block; and according to the position codes of the blood vessel segmentation images, filling each blood vessel segmentation image to a position corresponding to the position code to obtain an initial blood vessel image.

In one embodiment, extracting, according to the initial blood vessel image, a first image block within a preset range around a blood vessel centerline endpoint includes: extracting a blood vessel central line according to the initial blood vessel image; searching the end point of the blood vessel center line according to the blood vessel center line; and taking the image in a preset range around the center line end of the blood vessel as a first image block.

In one embodiment, the inputting the first image block into a blood vessel tracking network for performing endpoint tracking to obtain a final blood vessel image includes: inputting the first image block into a blood vessel tracking network for end point tracking to obtain an end point tracking image; extracting a connected domain according to the endpoint tracking image; and splicing the connected domain to the initial blood vessel image to obtain a final blood vessel image.

In one embodiment, the stitching the connected component to the initial blood vessel image to obtain a final blood vessel image includes: splicing the connected domain to an initial blood vessel image, and counting the splicing times; if the splicing times are more than or equal to the preset times, obtaining a final blood vessel image; and if the splicing times are less than the preset times, extracting the first image block in the preset range around the center line end of the blood vessel again for end point tracking until the splicing times are more than or equal to the preset times, and obtaining the final blood vessel image.

In one embodiment, the stitching the connected component to the initial blood vessel image to obtain a final blood vessel image includes: splicing the connected domain to the initial blood vessel image, and acquiring a first length of a blood vessel center line in an end point tracking image and a second length of the blood vessel center line in a first image block corresponding to the end point tracking image; the first length is different from the second length to obtain an increase distance; if the increasing distance is smaller than the preset distance, obtaining a final blood vessel image; if the increasing distance is larger than or equal to the preset distance, extracting the first image block in the preset range around the center line end point of the blood vessel again for blood vessel tracking until the increasing distance is smaller than the preset distance, and obtaining a final blood vessel image.

A vessel segmentation device comprising: the acquisition module is used for acquiring an image to be detected; the tissue organ segmentation module is used for inputting the image to be detected into a tissue organ segmentation network to obtain a tissue organ image; a distance field calculation module for obtaining a distance field of a blood vessel and a tissue organ corresponding to the blood vessel according to the tissue organ image; an annular band computation module for determining an annular band comprising the blood vessel based on the distance field and the tissue organ image; the blood vessel segmentation module is used for inputting the annular band into the blood vessel segmentation network to obtain an initial blood vessel image; the end point extraction module is used for extracting a first image block in a preset range around the end point of the central line of the blood vessel according to the initial blood vessel image; and the end point tracking module is used for inputting the first image block into a blood vessel tracking network for end point tracking to obtain a final blood vessel image.

A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the vessel segmentation method as any one of the above when executing the computer program.

A computer-readable storage medium, having stored thereon a computer program which, when executed by a processor, implements a vessel segmentation method as described in any of the above.

Compared with the related art, the blood vessel segmentation method provided by the embodiment of the application obtains the image to be detected, inputs the image to be detected into the tissue and organ segmentation network to obtain the tissue and organ image, obtains the distance field between the blood vessel and the tissue and organ corresponding to the blood vessel according to the tissue and organ image, determines the annular band including the blood vessel according to the distance field and the tissue and organ image, inputs the annular band into the blood vessel segmentation network to obtain the initial blood vessel image, extracts the first image block in the preset range around the central line end point of the blood vessel according to the initial blood vessel image, and inputs the first image block into the blood vessel tracking network to perform end point tracking to obtain the final blood vessel image. By the method, the accuracy and the stability of the acquired blood vessel image can be improved, the image is corrected through the neural network, the time and the energy of a technician are greatly saved, and the diagnosis cost is further saved.

The details of one or more embodiments of the application are set forth in the accompanying drawings and the description below to provide a more thorough understanding of the application.

Drawings

FIG. 1 is a flow diagram illustrating a method for vessel segmentation in one embodiment;

FIG. 2 is a schematic diagram of a network architecture in one embodiment;

FIG. 3 is a schematic diagram of a network architecture in another embodiment;

FIG. 4 is a flow diagram illustrating a method for obtaining an initial blood vessel image according to one embodiment;

FIG. 5 is a flow diagram illustrating a method for obtaining a final blood vessel image according to one embodiment;

FIG. 6 is a schematic flow chart of a method for obtaining a final blood vessel image according to another embodiment;

FIG. 7 is a process flow of coronary segmentation in one embodiment;

FIG. 8 is an image of a heart segmentation in one embodiment;

FIG. 9 is a schematic view of an endless belt in one embodiment;

FIG. 10(a) is a diagram of the effect of initial coronary segmentation, FIG. 10(b) is a diagram of the effect of coronary tracking, and FIG. 10(c) is a diagram of the gold standard for coronary segmentation;

FIG. 11(a) is an input image block of a coronary segmentation network; FIG. 11(b) is a diagram of the initial coronary segmentation result output by the coronary segmentation network; FIG. 11(c) is the coronary centerline extracted based on FIG. 11 (b); FIG. 11(d) is an input coronary endpoint image block of a coronary tracking network; FIG. 11(e) is a coronary tracking image output by the coronary tracking network; FIG. 11(f) is an image resulting from stitching a coronary tracking image to an initial coronary segmentation result;

FIG. 12 is a diagram illustrating a comparison of coronary segmentation effects in one embodiment;

FIG. 13 is a graphical illustration of a comparison of coronary segmentation results to gold standards in one embodiment;

FIG. 14 is a block diagram showing the structure of a blood vessel segmentation apparatus according to an embodiment;

FIG. 15 is a diagram showing an internal structure of a computer device according to an embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments provided in the present application without any inventive step are within the scope of protection of the present application.

It is obvious that the drawings in the following description are only examples or embodiments of the present application, and that it is also possible for a person skilled in the art to apply the present application to other similar contexts on the basis of these drawings without inventive effort. Moreover, it should be appreciated that in the development of any such actual implementation, as in any engineering or design project, numerous implementation-specific decisions must be made to achieve the developers' specific goals, such as compliance with system-related and business-related constraints, which may vary from one implementation to another.

Reference in the specification to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the specification. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. Those of ordinary skill in the art will explicitly and implicitly appreciate that the embodiments described herein may be combined with other embodiments without conflict.

Unless defined otherwise, technical or scientific terms referred to herein shall have the ordinary meaning as understood by those of ordinary skill in the art to which this application belongs. Reference to "a," "an," "the," and similar words throughout this application are not to be construed as limiting in number, and may refer to the singular or the plural. The present application is directed to the use of the terms "including," "comprising," "having," and any variations thereof, which are intended to cover non-exclusive inclusions; for example, a process, method, system, article, or apparatus that comprises a list of steps or modules (elements) is not limited to the listed steps or elements, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus. Reference to "connected," "coupled," and the like in this application is not intended to be limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The term "plurality" as referred to herein means two or more. "and/or" describes an association relationship of associated objects, meaning that three relationships may exist, for example, "A and/or B" may mean: a exists alone, A and B exist simultaneously, and B exists alone. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship. Reference herein to the terms "first," "second," "third," and the like, are merely to distinguish similar objects and do not denote a particular ordering for the objects.

In order to acquire an image to be detected of a scanning object, the scanning object needs to be scanned by a medical imaging device, wherein the scanning object may be a whole body organ of a patient, or an organ, a tissue, or a cell set, etc. that the patient needs to detect intensively. The medical imaging equipment scans a scanning object to obtain scanning data, and generates a medical image sequence according to the scanning data. Wherein the medical image sequence is an image of each cross section of the scanned object in the scanning direction. And finally generating a three-dimensional image of the internal structure of the scanned object according to the image sequence. Wherein the medical imaging device may be: x-ray imaging equipment, CT (normal CT, spiral CT), Positron Emission Tomography (PET), magnetic resonance imaging (MR), infrared scanning equipment, and combined scanning equipment of various scanning equipments, etc.

The following is exemplified with the medical imaging device as CT: a Computed Tomography (CT) apparatus typically includes a gantry, a couch, and a console for operation by a physician. One side of the frame is provided with a bulb tube, and the side opposite to the bulb tube is provided with a detector. The console is a computer device for controlling the bulb tube and the detector to scan, and the computer device is also used for receiving data collected by the detector, processing and reconstructing the data and finally forming a CT image. When CT is used for scanning, a patient lies on a scanning bed, the scanning bed sends the patient into the aperture of a stand, a bulb tube arranged on the stand emits X rays, the X rays penetrate through the patient and are received by a detector to form data, the data are transmitted to computer equipment, and the computer equipment carries out primary processing and image reconstruction on the data to obtain a CT image.

The embodiment also provides a blood vessel segmentation method. Fig. 1 is a schematic flow chart of a blood vessel segmentation method in an embodiment, as shown in fig. 1, the flow chart includes the following steps:

and step S101, acquiring an image to be detected.

Specifically, the medical imaging device scans a human body to be scanned first to obtain an image to be measured. The acquired image to be detected can be directly scanned by medical imaging equipment and directly reconstructed to obtain an image to be detected; the image to be detected can be obtained by scanning a human body to be scanned by medical imaging equipment, storing the scanning data, obtaining the stored scanning data and reconstructing to obtain the image to be detected; the image to be detected can also be obtained by scanning a human body to be scanned by medical imaging equipment and reconstructing the human body to be scanned, and the image to be detected is stored to obtain a stored image to be detected. The image to be detected is an image obtained by performing image reconstruction on the scanning data.

And S102, inputting the image to be detected into a tissue and organ segmentation network to obtain a tissue and organ image.

Specifically, the tissue and organ segmentation network adopts two neural networks from low resolution to high resolution through a deep learning method. And inputting the acquired image to be detected into a tissue and organ segmentation network to obtain a tissue and organ image, wherein the tissue and organ images such as stomach, lung, muscle and the like can be obviously and respectively obtained from the image.

Step S103, obtaining a distance field of a blood vessel and the tissue organ corresponding to the blood vessel according to the tissue organ image.

Specifically, a blood vessel exists in or near a tissue organ, the blood vessel is found in the tissue organ image, if the blood vessel is outside the tissue organ, namely near the tissue organ, the minimum distance from all pixel points of the tissue organ wall to the current blood vessel point is calculated, and all blood vessel points are traversed to obtain the extraluminal distance; if the blood vessel is in the tissue organ, namely in the tissue organ, the minimum distance from all pixel points of the tissue organ wall to the current blood vessel point is calculated, and all blood vessel points are traversed to obtain the intracavity distance. Wherein the distance field includes an extraluminal distance of the blood vessel outside the tissue organ and an intraluminal distance of the blood vessel within the tissue organ. Distance fields are used to represent the distance of a blood vessel to the wall of a tissue organ.

Step S104, determining an annular band containing the blood vessel according to the distance field and the tissue organ image.

Specifically, according to the intracavity distance and the extracavity distance of the distance field, an annular zone where a blood vessel is located is determined in the tissue organ image, the farthest distance from the wall of the tissue organ when the blood vessel is located outside the tissue organ is determined through the extracavity distance, and the farthest distance from the wall of the tissue organ when the blood vessel is located inside the tissue organ is determined through the intracavity distance, so that an annular zone where the blood vessel is located, namely the annular zone, is determined. The probability of the appearance of blood vessels inside the annulus is very high, while the probability of the appearance of blood vessels outside the annulus is very low.

And step S105, inputting the annular band into the blood vessel segmentation network to obtain an initial blood vessel image.

Specifically, the vessel segmentation network is a neural network obtained by a deep learning method, the vessel segmentation network comprises 4 upsampling and 4 downsampling, and comprises a feature extraction process of 5 scales, wherein the number of residual modules can be selected from 1 to 5. The input of the blood vessel segmentation network is a gray image in the annular band, namely a tissue organ image in the annular band, and the output is an initial blood vessel image. The initial blood vessel image is an image in which blood vessels are displayed.

And S106, extracting a first image block in a preset range around the center line end of the blood vessel according to the initial blood vessel image.

Specifically, a blood vessel central line is extracted according to the initial blood vessel image; searching the end point of the blood vessel center line according to the blood vessel center line; and taking the image in a preset range around the center line end of the blood vessel as a first image block. More specifically, according to the initial blood vessel image, a blood vessel centerline in the initial blood vessel image is first extracted, and a plurality of end points of the blood vessel centerline are found. For each end point, firstly, an image in a preset range around the end point is extracted as a first image block. Wherein the preset range may be an image within 5mm around the end point as the first image block. The preset range may be set according to a scene requirement in actual use, and this embodiment is not particularly limited. The central line of the blood vessel is a line formed by connecting the middle points of the blood vessel in the initial blood vessel image.

And S107, inputting the first image block into a blood vessel tracking network for end point tracking to obtain a final blood vessel image.

Specifically, the blood vessel tracking network is a neural network obtained by a deep learning method, and includes 2 times of upsampling and 2 times of downsampling, and emphasizes feature extraction of a local image. Inputting the first image block including the end point into a blood vessel tracking network for end point tracking to obtain an end point tracking image, extracting a connected domain from the end point tracking image, splicing the connected domain to the initial blood vessel image, and splicing the newly tracked pixel points to the initial blood vessel image to obtain a final blood vessel image. On the basis of carrying out the end point tracking once, the end point tracking and the splicing can be carried out for multiple times to obtain a final image.

The blood vessel segmentation method provided by this embodiment obtains a tissue organ image through a tissue organ segmentation network, determines a distance field through the tissue organ image, determines an annular zone where a blood vessel is located according to the distance field and the tissue organ image, determines an initial blood vessel image in the annular zone through the blood vessel segmentation network, and tracks end points of all blood vessels in the initial blood vessel image through a blood vessel tracking network to obtain a final blood vessel image. By the method, the accuracy and the stability of the acquired blood vessel image can be improved, the image is corrected through the neural network, the time and the energy of a technician are greatly saved, and the diagnosis cost is further saved.

In one embodiment, the tissue and organ segmentation network, the vessel segmentation network, and the vessel tracking network may be constructed based on: convolutional Neural Networks (CNN), Generative Antagonistic Networks (GAN), or the like, or combinations thereof. Examples of Convolutional Neural Networks (CNNs) may include SRCNN (Super-Resolution Convolutional Neural Network), DnCNN (Denoising Convolutional Neural Network), U-net, V-net, and FCN (full Convolutional Neural Network). In some embodiments, the neural network model may include multiple layers, such as an input layer, multiple hidden layers, and an output layer. The plurality of hidden layers may include one or more convolutional layers, one or more bulk normalization layers, one or more active layers, fully-connected layers, cost function layers, and the like. Each of the plurality of layers may include a plurality of nodes.

In one embodiment, the tissue organ segmentation network and the blood vessel segmentation network may be network structures as shown in fig. 2. Specifically, the network structure comprises 4 times of down sampling and 4 times of up sampling, and comprises a feature extraction process with 5 scales, wherein the number of residual modules can be selected from 1 to 5. During the training process of the neural network, the size of the cut image block can be 64-256 pixels, and the preferred size is 128 × 128. The resolution of the image cut by the neural network may be 0.3-0.8mm, preferably 0.6 mm. When the neural network normalizes the image, the normalization is performed by adopting a mean value and a standard deviation, wherein the mean value is preferably 50, and the standard deviation is preferably 400. The neural network adopts minimization of dice-loss coefficient as an optimization target. The neural network adopts an Adam parameter optimization method.

In one embodiment, the vessel tracking network may be a network structure as shown in fig. 3. Specifically, the network structure includes 2 upsampling and 2 downsampling, and focuses on feature extraction of local images. The size of the image blocks cut out by the neural network during training can be 8-32 pixels, and the preferred size is 16 × 16. The neural network cuts images with a resolution of 0.3-0.8mm, preferably 0.6 mm. When the neural network normalizes the image, the normalization is performed by adopting a mean value and a standard deviation, wherein the mean value is preferably 50, and the standard deviation is preferably 400. The neural network adopts minimization of dice-loss coefficient as an optimization target. The neural network adopts an Adam parameter optimization method. The sampling mode of the neural network is to sample positive samples on blood vessels and sample negative samples in 5mm areas outside the blood vessels. Through the network structure shown in fig. 3, the conventional network structure sets multiple times of sampling, while the present embodiment only sets 2 times of upsampling and 2 times of downsampling, so that the calculation amount of sampling is reduced, the feature extraction of the local image can be satisfied, and the accuracy of the local feature extraction can be ensured.

In one embodiment, a method for acquiring an initial blood vessel image is provided, and fig. 4 is a schematic flowchart of the method for acquiring an initial blood vessel image in one embodiment, as shown in fig. 4, the flowchart includes the following steps:

step S401: and dividing the annular band into a plurality of second image blocks according to a preset size.

Specifically, a point is selected on the annular band, and the second image block is sequentially and continuously intercepted by taking the point as a starting point and a preset step length and a block size. Wherein step size represents how far to sample a cube; the tile size refers to the size of the tile to be cut. The step size needs to be smaller than or equal to the size of the cut-out so that the image in the whole annular band is cut out and partial areas are prevented from being uncut. The smaller the step size is, the denser the sampling density of the image blocks is, the larger the overlapping area among the cut image blocks is, and the slower the cutting speed of the whole annular belt is. Preferably, the step size is consistent with the tile size, i.e. there is no overlap between two adjacent image blocks. Such a cutting method can maximize the cutting speed in the case where all the images in the annular band are cut. The minimum preset step length is 10mm, and the example is that the preset step length is 10mm, and the size of the cut block is 10 mm: intercepting is carried out from the upper left corner of a bounding box (a bounding box) of the annular band until the lower right corner is finished, an image block with the length of 10mm is intercepted every interval of 10mm, an image with the image length of 10mm is intercepted on the left side or the right side of the starting point to serve as a first second image block, then a second image block is intercepted at the starting point along the direction of intercepting the first second image block by 10mm, and a plurality of second image blocks are intercepted continuously and sequentially without intervals. Wherein, bounding box is a bounding box and represents the largest cube outside an object. Preferably, a first second image block is truncated at a first position (10, 10, 10 to 41, 41, 41), the size of the image block is 32 × 32, and then a second image block is truncated at a next position (42, 42, 42 to 74, 74, 74) of the first second image block until the truncation is completed.

Step S402: and respectively inputting the second image blocks into the blood vessel segmentation network to obtain a plurality of blood vessel segmentation images.

Specifically, a plurality of captured second image blocks are respectively input to the blood vessel segmentation network, and each second image block is corresponding to one obtained blood vessel segmentation image.

Step S403: and splicing the plurality of blood vessel segmentation images to obtain an initial blood vessel image.

Specifically, the obtained multiple blood vessel segmentation images are spliced in sequence to obtain an initial blood vessel image. The second image block comprises a position code. Specifically, the annular band is divided, and when a plurality of second image blocks are sequentially and continuously cut at the starting position at a preset step size without intervals, a position code is allocated to each second image block, and the position code can identify the position of the second image block in the annular band. Acquiring the position code of each blood vessel segmentation image in an annular band according to the position code of each second image block; and according to the position codes of the blood vessel segmentation images, filling each blood vessel segmentation image to a position corresponding to the position code to obtain an initial blood vessel image. More specifically, each second image block comprises a position code, after the second image blocks are input into the blood vessel segmentation network, the position codes of the second image blocks are distributed to corresponding blood vessel segmentation images, and all the blood vessel segmentation images are filled to the positions corresponding to the position codes according to the positions identified by the position codes to obtain the initial blood vessel image.

In one of the embodiments, the input of the vessel segmentation network may not be limited to an image patch cut by an annular band. The image block cut by the annular band can be used as a first input channel, the tissue and organ image within the position range of the image block cut by the annular band can be used as a second input channel, and the first channel image and the second channel image are input into the blood vessel segmentation network.

According to the method for acquiring the initial blood vessel image, the blood vessel image corresponding to the tissue organ is acquired on the basis of tissue organ segmentation. Because the structure of the blood vessel is more complicated and changeable, and the tissue organ is more stable, therefore, firstly, a deep learning model is adopted, the image is firstly segmented into the images of all the tissue organs, reliable annular positioning is provided for the segmentation of the blood vessel through the images of the tissue organs, a large number of regions with smaller probability of the blood vessel are removed, the training efficiency and the convergence rate of a blood vessel segmentation network can be improved, and meanwhile, the possible errors of the blood vessel segmentation can be greatly reduced.

In one embodiment, a method for acquiring a final blood vessel image is provided, and fig. 5 is a schematic flowchart of the method for acquiring a final blood vessel image in one embodiment, as shown in fig. 5, the flowchart includes the following steps:

step S501: and inputting the first image block into a blood vessel tracking network for end point tracking to obtain an end point tracking image.

Specifically, the endpoint-tracking image is an image obtained by tracking and extending the endpoint on the basis of the endpoint of the first image block.

Step S502: and extracting a connected domain according to the endpoint tracking image.

Specifically, the connected component generally refers to an image region composed of foreground pixels having the same pixel value and adjacent positions in the image. In the endpoint tracking image, the area is the area formed by the pixel points where the blood vessels are located. That is, the connected domain where the blood vessel is located is extracted in the endpoint tracking image.

Step S503: and splicing the connected domain to an initial blood vessel image, and counting the splicing times.

Specifically, the connected domain acquired in the above steps is spliced to the corresponding endpoint of the initial blood vessel image, and the number of times of splicing the connected domain at the endpoint is counted.

Step S504: and if the splicing times are more than or equal to the preset times, obtaining a final blood vessel image.

Step S505: and if the splicing times are less than the preset times, extracting the first image block in the preset range around the center line end of the blood vessel again for end point tracking until the splicing times are more than or equal to the preset times, and obtaining the final blood vessel image.

Specifically, the preset number of times may be 3 to 20 times. For example, the preset number of times is 5, and if the number of times of splicing the connected domain at the end point is greater than or equal to 5 times, an image obtained after the last splicing is completed is used as a final blood vessel image. If the number of times of splicing the connected domain at the end point is less than 5 times, acquiring the end point of the center line of the blood vessel again based on the blood vessel image spliced at the last time, selecting an image in a preset range around the end point as a first image block, and inputting the first image block into a blood vessel tracking network for end point tracking to obtain an end point tracking image; extracting a connected domain according to the endpoint tracking image; and splicing the connected domain to the blood vessel image which is spliced at the last time, and counting the splicing times of the end points again until the splicing times are more than or equal to 5 times, thus obtaining the final blood vessel image.

In one embodiment, the input to the vessel tracking network is not limited to the first image block within a preset range around the end point. The first image block may be used as a first input channel, any one of the tissue and organ image blocks and the initial blood vessel image block within the position range of the first image block may be used as a second input channel, and both the first input channel and the second input channel may be input to the blood vessel tracking network. Or the first image block may be used as a first input channel, the tissue and organ image block within the position range of the first image block may be used as a second input channel, the initial blood vessel image block within the position range of the first image block may be used as a third channel, and the simultaneous input channels of the first input channel, the second input channel, and the third channel may be input to the blood vessel tracking network.

In one embodiment, the blood vessel tracking method may extract a blood vessel direction according to an initial segmentation result of the blood vessel before extracting an image block of a center line end point of the blood vessel, and then extract the blood vessel according to a main direction, so as to ensure that the direction of the blood vessel is parallel to an x direction (or a y direction and a z direction) of the extracted image block, thereby achieving the effects of simplifying depth learning and improving segmentation performance.

According to the method for obtaining the final blood vessel image, the segmentation model network based on the whole image is broken frequently, and the clinical practical use is influenced. And the vessel tracking model network based on the local image greatly simplifies the segmentation task, and only needs to distinguish the enhanced vessel from the non-enhanced vessel. Our images to be measured are CTA data, i.e. data that have been contrast agent-injected; the coronary vessels are all brighter vessels at this time. The vessel segmentation network focuses on the overall vessel segmentation, i.e. separating the coronary vessels from other tissues. Other tissues include other non-coronary vessels and other tissue organs including, for example, the heart, lungs, ribs, etc. The task of the blood vessel tracking network is similar to that of the blood vessel segmentation network, coronary blood vessels and other tissues are separated, only the action range is greatly reduced, the coronary blood vessels and other tissues are fracture tissues of an initial blood vessel image, the content and complexity of the tissues to be identified are greatly simplified, and the blood vessel tissues and the tissues which are very adjacent to the blood vessel tissues are generally separated. Wherein the very adjacent tissue is typically only non-enhanced blood vessels, a portion of lung tissue, and a portion of heart tissue. Therefore, the segmentation of the broken blood vessels can be effectively connected, and the integrity of the segmentation of the blood vessels is improved. And through setting up and predetermine the concatenation number of times, can improve and cut apart efficiency under the prerequisite of guaranteeing to cut apart the blood vessel image accuracy.

In one embodiment, a method for acquiring a final blood vessel image is provided, and fig. 6 is a schematic flow chart of the method for acquiring a final blood vessel image in another embodiment, as shown in fig. 6, the flow chart includes the following steps:

step S601: and inputting the first image block into a blood vessel tracking network for end point tracking to obtain an end point tracking image.

Step S602: and extracting a connected domain according to the endpoint tracking image.

Step S603: and splicing the connected domain to the initial blood vessel image, and acquiring a first length of a blood vessel center line in an end point tracking image and a second length of the blood vessel center line in a first image block corresponding to the end point tracking image.

Specifically, the connected domain acquired in the above step is spliced to the corresponding end point of the initial blood vessel image, the first length of the blood vessel center line in the end point tracking image corresponding to the spliced connected domain is acquired, and the second length of the blood vessel center line in the first image block corresponding to the end point tracking image is acquired.

Step S604: and obtaining the increasing distance by subtracting the first length from the second length.

Step S605: and if the increasing distance is smaller than the preset distance, obtaining a final blood vessel image.

Step S606: if the increasing distance is larger than or equal to the preset distance, extracting the first image block in the preset range around the center line end point of the blood vessel again for blood vessel tracking until the increasing distance is smaller than the preset distance, and obtaining a final blood vessel image.

Specifically, the distance description is performed by taking the preset distance as 1mm as an example, and if the increase distance is less than 1mm, an image obtained after the last stitching is completed is taken as a final blood vessel image. If the increase distance is larger than or equal to 1mm, acquiring the end point of the center line of the blood vessel again based on the blood vessel image spliced at the previous time, selecting an image in a preset range around the end point as a first image block, and inputting the first image block into a blood vessel tracking network for end point tracking to obtain an end point tracking image; extracting a connected domain according to the endpoint tracking image; and splicing the connected domain to the blood vessel image which is spliced at the previous time, and calculating the increase distance again until the increase distance is less than 1mm to obtain the final blood vessel image.

According to the method for obtaining the final blood vessel image, the segmentation model network based on the whole image is broken frequently, and the clinical practical use is influenced. The vessel tracking model network based on the local image greatly simplifies the segmentation task, and only needs to distinguish the enhanced vessel from the non-enhanced vessel, so that the segmentation of the broken vessel can be effectively connected, and the integrity of the vessel segmentation is improved, namely the vessel segmentation network and the vessel tracking network are in a serial relationship, the vessel is segmented first, and then the vessel tracking is carried out. And setting a preset distance, and when the increasing distance is smaller than the preset distance, indicating that the increasing distance of the current endpoint is about to reach the limit. The tracking condition of the end point can be accurately judged by setting a proper preset distance, and the blood vessel tracking can be stopped timely, so that the tracking effect of the end point can be ensured, the tracking time can be saved, and the segmentation efficiency can be improved.

In one embodiment, the number of splices and the increase distance may be detected simultaneously, and when any one of the conditions is satisfied, the end point tracking is ended to obtain a final blood vessel image.

In one embodiment, the medical imaging device is a CT and the blood vessel is a coronary artery. As shown in fig. 7, a coronary artery segmentation process is provided, in which an image to be measured obtained by CT scanning is first obtained, a heart segmentation based on deep learning is first performed on the image to be measured, an annular band is generated by calculating a distance field on the basis of the heart segmentation, a coronary artery rough segmentation based on the deep learning is performed on the annular band to obtain an initial coronary artery image, and then a coronary artery tracking based on the deep learning is performed on the basis of the initial coronary artery image, so that the more accurate and stable coronary artery segmentation effect is obtained by combining the stability of the heart segmentation, the integrity of the initial coronary artery segmentation and the locality of the coronary artery tracking.

More specifically, an object to be scanned is scanned by a Computed Tomography (CT) device to obtain original CT data, and an image to be measured is generated. Inputting the image to be measured into a heart segmentation network, and performing heart segmentation to obtain a heart segmentation image shown in fig. 8, wherein the heart segmentation image includes: aorta, right atrium, right ventricle, left atrium, and left ventricle. The statistical table of the position relationship between the coronary artery and the heart segmentation result is obtained by performing distance statistics of the heart and the coronary artery range on 800 heart segmentation images, and is as follows:

TABLE 1 statistics of the positional relationship between coronary arteries and cardiac segmentation results

According to table 1, we perform distance field calculation from the foreground and from the background on the heart segmentation result, and obtain the region where the coronary artery may appear, i.e., the annular band, as shown in fig. 9, where the black annular region is the region where the coronary artery may appear, i.e., the annular band, the blood vessels of various colors in the black annular band are the branches of the coronary artery, and the other regions are the regions with very low probability of appearance of the coronary artery.

And cutting effective image blocks in the annular band, namely, sequentially and continuously cutting a plurality of image blocks without intervals by a fixed step length, and sending the image blocks into a coronary artery segmentation network to obtain an initial coronary artery segmentation result. As shown in fig. 10, fig. 10(a) shows the initial coronary artery segmentation effect, fig. 10(b) shows the coronary artery tracking effect, and fig. 10(c) shows the coronary artery segmentation golden standard. As can be seen from fig. 10, many segmented and fractured regions appear in the image with the initial coronary artery segmentation effect, and on this basis, the initial coronary artery segmentation effect is followed by coronary artery tracking through the coronary artery tracking network to obtain an accurate coronary artery segmentation result.

In one embodiment, a coronary tracking process and effect is performed, as shown in FIG. 11, where FIG. 11(a) is an input image block of a coronary segmentation network; FIG. 11(b) shows the initial coronary segmentation result output by the coronary segmentation network; FIG. 11(c) is the coronary centerline extracted based on FIG. 11 (b); FIG. 11(d) is an input coronary endpoint image block of a coronary tracking network; FIG. 11(e) is a coronary tracking image output by the coronary tracking network; fig. 11(f) is an image obtained by stitching a coronary tracking image to the initial coronary segmentation result. Fig. 11(a) is a three-view of an image block, and because the selected slice size is large, one slice contains all the coronaries. Fig. 11(b) is an initial coronary image, fig. 11(f) is a final coronary image obtained after coronary tracking, and fig. 11(f) is a diagram showing a coronary growth at the position of the end point of the coronary artery and the position of the end point of the fracture with respect to fig. 11 (b). First, the coronary artery centerline is extracted from the initial coronary artery segmentation result outputted from the coronary artery segmentation network, and the extraction result of the centerline is shown in fig. 11 (c). The end point is extracted on the coronary centerline, that is, the point where the neighborhood has only 1 is extracted, as shown by the intersection point position of the dotted line in fig. 11 (c). Taking each end point as a center, cutting image blocks respectively, as shown in fig. 11(d), and sending the image blocks into a trained coronary artery tracking network to obtain a coronary artery tracking image, as shown in fig. 11 (e). A connected domain including the midpoint is extracted from the coronary tracking image, and the connected domain is then stitched to the initial coronary segmentation result, resulting in the image shown in fig. 11 (f). The connected domain generally refers to an image region formed by foreground pixels with the same pixel value and adjacent positions in an image. In the endpoint tracking image, the area is the area formed by the pixel points where the blood vessels are located. The midpoint is also the midpoint of the vessel. Finally, judging the coronary artery segmentation result, judging whether the end point of the coronary artery needs to be tracked continuously, and stopping tracking when the following optional conditions are met: under the condition one, the tracking times are more than or equal to N1, and the selectable range of N1 is 3-20 times; and secondly, extracting a central line from the obtained connected domain, comparing the central line with the original segmentation result or the central line, wherein the growth distance of the central line is less than N2, and N2 can be 1/2 cut block image sizes. If the condition is met, obtaining a final coronary image; and if not, continuing to track the coronary artery, and finally meeting the conditions to obtain a final coronary artery image. The above embodiments, coronary segmentation techniques based on the results of cardiac segmentation. Compared with the complexity and changeability of coronary vessel structure, the atrium and ventricle structure of the heart is more stable. Therefore, each chamber of the heart is firstly segmented by adopting a deep learning model, reliable annular positioning of a coronary artery range is provided for coronary artery segmentation, and a large number of possible non-coronary artery regions are removed. The training efficiency and the convergence speed of the coronary artery segmentation model can be improved, and errors possibly caused by coronary artery segmentation are greatly reduced.

The embodiment is a method for combining a whole image-based blood vessel segmentation network and a local image-based blood vessel tracking network. The whole image blood vessel segmentation network has a larger image visual field, can effectively distinguish coronary arteries and veins adjacent to the coronary arteries enhanced by the imaging agent by utilizing the heart structure, and reduces false positives (namely segmenting blood vessels outside the coronary arteries). However, due to the complexity and high variability of coronary vessel structures, the vessels are slender and the image resolution is limited, the segmentation result output based on the vessel segmentation network often forms fractures, which affects the practical clinical use. The vessel tracking network based on the local image greatly simplifies the segmentation task, and only needs to distinguish the enhanced vessel from the non-enhanced vessel, so that the segmentation of the broken vessel can be effectively connected, and the integrity of the vessel segmentation is improved. The invention can more accurately and completely segment the coronary artery by adopting the method of combining the blood vessel segmentation network based on the whole image and the blood vessel tracking network of the local image.

The above embodiment was tested by collecting 189 CTA chest data not used for network training and asking the doctor to label the coronary vessels (labeling of doctor is strictly in accordance with the standard of SCCT18 branch, SCCT: American society for cardiovascular CT), and the coronary vessel segmentation method of the above embodiment was used to segment the coronary vessels, and then the comparison of the results is performed, as shown in FIG. 12, wherein the comparison of the results of the coronary vessel segmentation method of the above embodiment with the manually labeled segmentation results after passing through the vessel segmentation network is shown in the first column of FIG. 12, and the comparison of the vessel segmentation results after adding the vessel tracking with the manually labeled results is shown in the second column of the following figure, wherein DICE is a measurement based on the whole segmentation results, and OV is a measurement based on the whole segmentation results^[4]The coronary vessel is a slender tubular structure, branches in different areas have thickness radius difference of several times, and the measurement standard based on the vessel center line is adopted, so that errors such as fracture and the like which often occur in vessel segmentation can be reflected more sensitively and more accurately compared with a DICE coefficient based on the whole vessel segmentation result. In addition, the judgment of coronary artery focus by the clinical imaging doctor mostly depends on the visualization technology such as CPR, S-CPR or cross section extracted based on the coronary artery central line, so the segmentation measurement index OV based on the central line can more accurately reflect the requirement of practical clinical application.

Wherein the minimum distance from the center line point of the gold standard to the center line of the segmentation result is less than the number of the radii of the segmentation resultReferred to as TPR_OVThe number of points larger than the radius of the division result is called FN_OVThe number of the points of which the distance from the center line point of the segmentation result to the center line of the golden standard result is less than the radius of the golden standard is called TPM_OVThe number of points larger than the radius of the gold standard is called FP_OV。

As can be seen in fig. 12: 1. after vessel tracking, the overall performance (DICE and OV) was similar; 2. after the blood vessel tracking, the recall value is greatly improved, and the repairing effect of the blood vessel tracking on the broken blood vessel is reflected; 3. after vessel tracking, the precision value is reduced because the vessel tracking network traces back to the end of a deeper and thinner vessel or a coronary branch not in the SCCT18 segment, such as the segmentation result shown in fig. 13, where fig. 13 is a diagram comparing the coronary segmentation result with the gold standard in one embodiment; the left image is the coronary artery segmentation result of the embodiment; the right image is marked by a gold standard, different line segments represent different coronary branches, and the segmentation part indicated by the arrow is OM3 which is not in the range of SCCT18 segment but belongs to coronary blood vessels, thus having clinical value as well.

It should be understood that although the various steps in the flowcharts of fig. 1, 4-6 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least some of the steps in fig. 1, 4-6 may include multiple sub-steps or multiple stages that are not necessarily performed at the same time, but may be performed at different times, and the order of performance of the sub-steps or stages is not necessarily sequential, but may be performed in turn or alternating with other steps or at least some of the sub-steps or stages of other steps.

In one embodiment, as shown in fig. 14, there is provided a blood vessel segmentation device, including: an acquisition module 100, a tissue organ segmentation module 200, a distance field calculation module 300, an annular band calculation module 400, a vessel segmentation module 500, an end point extraction module 600, and an end point tracking module 700, wherein:

an obtaining module 100, configured to obtain an image to be detected;

a tissue and organ segmentation module 200, configured to input the image to be detected into a tissue and organ segmentation network, so as to obtain a tissue and organ image;

a distance field calculation module 300, configured to obtain a distance field of a blood vessel and a tissue organ corresponding to the blood vessel according to the tissue organ image;

an annular band computation module 400 for determining an annular band comprising the vessel based on the distance field and the tissue organ image;

a blood vessel segmentation module 500, configured to input the annular band into the blood vessel segmentation network, so as to obtain an initial blood vessel image;

an endpoint extraction module 600, configured to extract, according to the initial blood vessel image, a first image block within a preset range around a blood vessel centerline endpoint;

the endpoint tracking module 700 is configured to input the first image block into a blood vessel tracking network for endpoint tracking, so as to obtain a final blood vessel image.

The blood vessel segmentation module 500 is further configured to segment the annular band into a plurality of second image blocks according to a preset size; respectively inputting the second image blocks into the blood vessel segmentation network to obtain a plurality of blood vessel segmentation images; and splicing the plurality of blood vessel segmentation images to obtain an initial blood vessel image.

The blood vessel segmentation module 500 is further configured to obtain a position code of each blood vessel segmentation image in the annular band according to the position code of each second image block; and according to the position codes of the blood vessel segmentation images, filling each blood vessel segmentation image to a position corresponding to the position code to obtain an initial blood vessel image.

An endpoint extraction module 600, configured to extract a blood vessel centerline according to the initial blood vessel image; searching the end point of the blood vessel center line according to the blood vessel center line; and taking the image in a preset range around the center line end of the blood vessel as a first image block.

The endpoint tracking module 700 is further configured to input the first image block into a blood vessel tracking network for endpoint tracking, so as to obtain an endpoint tracking image; extracting a connected domain according to the endpoint tracking image; and splicing the connected domain to the initial blood vessel image to obtain a final blood vessel image.

The endpoint tracking module 700 is further configured to splice the connected domain to the initial blood vessel image, and count the number of times of splicing; if the splicing times are more than or equal to the preset times, obtaining a final blood vessel image; and if the splicing times are less than the preset times, extracting the first image block in the preset range around the center line end of the blood vessel again for end point tracking until the splicing times are more than or equal to the preset times, and obtaining the final blood vessel image.

The endpoint tracking module 700 is further configured to splice the connected domain to the initial blood vessel image, and obtain a first length of a blood vessel center line in the endpoint tracking image and a second length of the blood vessel center line in the first image block corresponding to the endpoint tracking image; the first length is different from the second length to obtain an increase distance; if the increasing distance is smaller than the preset distance, obtaining a final blood vessel image; if the increasing distance is larger than or equal to the preset distance, extracting the first image block in the preset range around the center line end point of the blood vessel again for blood vessel tracking until the increasing distance is smaller than the preset distance, and obtaining a final blood vessel image.

For specific definition of the blood vessel segmentation device, reference may be made to the above definition of the blood vessel segmentation method, which is not described herein again. The modules in the blood vessel segmentation device can be wholly or partially realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.

In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown in fig. 15. The computer device includes a processor, a memory, a network interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a vessel segmentation method. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.

Those skilled in the art will appreciate that the architecture shown in fig. 15 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.

In one embodiment, a computer device is provided, comprising a memory and a processor, the memory having a computer program stored therein, the processor implementing the following steps when executing the computer program:

acquiring an image to be detected; inputting the image to be detected into a tissue and organ segmentation network to obtain a tissue and organ image; obtaining a blood vessel and a distance field of the tissue organ corresponding to the blood vessel according to the tissue organ image; determining an annular band containing the blood vessel from the distance field and the tissue organ image; inputting the annular band into the blood vessel segmentation network to obtain an initial blood vessel image; extracting a first image block within a preset range around the center line end point of the blood vessel according to the initial blood vessel image; and inputting the first image block into a blood vessel tracking network for end point tracking to obtain a final blood vessel image.

In one embodiment, the processor, when executing the computer program, further performs the steps of:

dividing the annular band into a plurality of second image blocks according to a preset size; respectively inputting the second image blocks into the blood vessel segmentation network to obtain a plurality of blood vessel segmentation images; and splicing the plurality of blood vessel segmentation images to obtain an initial blood vessel image.

acquiring the position code of each blood vessel segmentation image in an annular band according to the position code of each second image block; and according to the position codes of the blood vessel segmentation images, filling each blood vessel segmentation image to a position corresponding to the position code to obtain an initial blood vessel image.

extracting a blood vessel central line according to the initial blood vessel image; searching the end point of the blood vessel center line according to the blood vessel center line; and taking the image in a preset range around the center line end of the blood vessel as a first image block.

inputting the first image block into a blood vessel tracking network for end point tracking to obtain an end point tracking image; extracting a connected domain according to the endpoint tracking image; and splicing the connected domain to the initial blood vessel image to obtain a final blood vessel image.

splicing the connected domain to an initial blood vessel image, and counting the splicing times; if the splicing times are more than or equal to the preset times, obtaining a final blood vessel image; and if the splicing times are less than the preset times, extracting the first image block in the preset range around the center line end of the blood vessel again for end point tracking until the splicing times are more than or equal to the preset times, and obtaining the final blood vessel image.

splicing the connected domain to the initial blood vessel image, and acquiring a first length of a blood vessel center line in an end point tracking image and a second length of the blood vessel center line in a first image block corresponding to the end point tracking image; the first length is different from the second length to obtain an increase distance; if the increasing distance is smaller than the preset distance, obtaining a final blood vessel image; if the increasing distance is larger than or equal to the preset distance, extracting the first image block in the preset range around the center line end point of the blood vessel again for blood vessel tracking until the increasing distance is smaller than the preset distance, and obtaining a final blood vessel image.

In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which when executed by a processor, performs the steps of:

In one embodiment, the computer program when executed by the processor further performs the steps of:

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.

The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims

1. A method of vessel segmentation, comprising:

acquiring an image to be detected;

inputting the image to be detected into a tissue and organ segmentation network to obtain a tissue and organ image;

obtaining a blood vessel and a distance field of the tissue organ corresponding to the blood vessel according to the tissue organ image;

determining an annular band containing the blood vessel from the distance field and the tissue organ image;

inputting the annular band into the blood vessel segmentation network to obtain an initial blood vessel image;

extracting a first image block within a preset range around the center line end point of the blood vessel according to the initial blood vessel image;

and inputting the first image block into a blood vessel tracking network for end point tracking to obtain a final blood vessel image.

2. The vessel segmentation method according to claim 1, wherein inputting the annular band into the vessel segmentation network, obtaining an initial vessel image comprises:

dividing the annular band into a plurality of second image blocks according to a preset size;

respectively inputting the second image blocks into the blood vessel segmentation network to obtain a plurality of blood vessel segmentation images;

and splicing the plurality of blood vessel segmentation images to obtain the initial blood vessel image.

3. The method of claim 2, wherein the stitching the vessel segmentation images to obtain the initial vessel image comprises: the second image block comprises a position code;

acquiring the position code of each blood vessel segmentation image in an annular band according to the position code of each second image block;

and according to the position codes of the blood vessel segmentation images, filling each blood vessel segmentation image to a position corresponding to the position code to obtain the initial blood vessel image.

4. The blood vessel segmentation method according to claim 1, wherein extracting, according to the initial blood vessel image, a first image block within a preset range around a center line end of a blood vessel comprises:

extracting a blood vessel central line according to the initial blood vessel image;

searching the end point of the blood vessel center line according to the blood vessel center line;

and taking the image in a preset range around the center line end of the blood vessel as a first image block.

5. The vessel segmentation method according to claim 1, wherein the inputting the first image block into a vessel tracking network for performing endpoint tracking to obtain a final vessel image comprises:

inputting the first image block into a blood vessel tracking network for end point tracking to obtain an end point tracking image;

extracting a connected domain according to the endpoint tracking image;

and splicing the connected domain to the initial blood vessel image to obtain the final blood vessel image.

6. The vessel segmentation method according to claim 5, wherein the stitching the connected component to the initial vessel image to obtain the final vessel image comprises:

splicing the connected domain to an initial blood vessel image, and counting the splicing times;

if the splicing times are more than or equal to the preset times, obtaining a final blood vessel image;

and if the splicing times are less than the preset times, extracting the first image blocks in the preset range around the center line end of the blood vessel again for end point tracking until the splicing times are more than or equal to the preset times, and obtaining the final blood vessel image.

7. The vessel segmentation method according to claim 5, wherein the stitching the connected component to the initial vessel image to obtain a final vessel image comprises:

splicing the connected domain to the initial blood vessel image, and acquiring a first length of a blood vessel center line in an end point tracking image and a second length of the blood vessel center line in a first image block corresponding to the end point tracking image;

the first length is different from the second length to obtain an increase distance;

if the increasing distance is smaller than the preset distance, obtaining a final blood vessel image;

if the increasing distance is larger than or equal to the preset distance, extracting the first image block in the preset range around the center line end point of the blood vessel again for blood vessel tracking until the increasing distance is smaller than the preset distance, and obtaining the final blood vessel image.

8. A vessel segmentation device, comprising:

the acquisition module is used for acquiring an image to be detected;

the tissue organ segmentation module is used for inputting the image to be detected into a tissue organ segmentation network to obtain a tissue organ image;

a distance field calculation module for obtaining a distance field of a blood vessel and a tissue organ corresponding to the blood vessel according to the tissue organ image;

an annular band computation module for determining an annular band comprising the blood vessel based on the distance field and the tissue organ image;

the blood vessel segmentation module is used for inputting the annular band into the blood vessel segmentation network to obtain an initial blood vessel image;

the end point extraction module is used for extracting a first image block in a preset range around the end point of the central line of the blood vessel according to the initial blood vessel image;

and the end point tracking module is used for inputting the first image block into a blood vessel tracking network for end point tracking to obtain a final blood vessel image.

9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the vessel segmentation method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out a vessel segmentation method as claimed in any one of claims 1 to 7.