WO2022217896A1 - 血管图像检测方法及检测模型训练方法、相关装置、设备 - Google Patents

血管图像检测方法及检测模型训练方法、相关装置、设备 Download PDF

Info

Publication number
WO2022217896A1
WO2022217896A1 PCT/CN2021/128160 CN2021128160W WO2022217896A1 WO 2022217896 A1 WO2022217896 A1 WO 2022217896A1 CN 2021128160 W CN2021128160 W CN 2021128160W WO 2022217896 A1 WO2022217896 A1 WO 2022217896A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
blood vessel
region
detection
dimensional
Prior art date
Application number
PCT/CN2021/128160
Other languages
English (en)
French (fr)
Inventor
李卓威
夏清
黄宁
Original Assignee
上海商汤智能科技有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 上海商汤智能科技有限公司 filed Critical 上海商汤智能科技有限公司
Publication of WO2022217896A1 publication Critical patent/WO2022217896A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/20—Image enhancement or restoration using local operators
    • G06T5/30—Erosion or dilatation, e.g. thinning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • 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
    • G06T2207/10016—Video; Image sequence
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30101—Blood vessel; Artery; Vein; Vascular

Definitions

  • the present application relates to the technical field of image processing, and in particular, to a blood vessel image detection method, a detection model training method, and related devices and equipment.
  • the present application provides a blood vessel image detection method, a detection model training method, and related devices and equipment.
  • a first aspect of the present application provides a blood vessel image detection method, comprising: acquiring a three-dimensional blood vessel image of an object to be measured; extracting at least one frame of two-dimensional blood vessel image from the three-dimensional blood vessel image along a direction perpendicular to the extending direction of the blood vessel in the three-dimensional blood vessel image image; the two-dimensional image is detected to obtain a region detection result of the two-dimensional image, wherein the region detection result includes a first sub-region corresponding to the plaque in the blood vessel.
  • the detection accuracy of plaque in the blood vessel is improved, and it is beneficial to provide data support for the subsequent quantification of the stenosis degree value of the blood vessel.
  • the area detection result further includes the first area corresponding to the blood vessel; after the two-dimensional image is detected and the area detection result of the two-dimensional image is obtained, the method further includes: based on the first area in at least one frame of the two-dimensional image and the In the first sub-region, the stenosis degree value of the blood vessel to be measured is obtained.
  • the area detection result also includes the first area corresponding to the blood vessel, so that the stenosis degree value of the blood vessel to be measured can be obtained based on the first area and the first sub-area in the at least one frame of two-dimensional image, thereby improving user experience.
  • obtaining the stenosis degree value of the blood vessel to be measured based on the first area and the first sub-area in the at least one frame of two-dimensional image includes: using the area of the first sub-area and the first area of the first area to determine the stenosis degree value; or, using the smallest first width of the first region in the radial direction and the smallest second width of the first sub-region in the radial direction of the first region to determine the stenosis degree value; or, in the three-dimensional blood vessel image
  • the stenosis degree value is determined by using the difference between the area of the first region and the area of the first sub-region in the multiple frames of two-dimensional images; or, when the three-dimensional blood vessel image includes multiple frames of two-dimensional images
  • the stenosis degree value is determined, or by using the smallest first width of the first region in the radial direction and the first width of the first subregion in the first region.
  • the smallest second width in the radial direction is used to determine the stenosis degree value, or when the three-dimensional blood vessel image includes multiple frames of two-dimensional images, the area of the first region and the area of the first sub-region in the multiple frames of two-dimensional images are used to determine the value of the stenosis degree.
  • the difference value is used to determine the stenosis degree value, or when the three-dimensional blood vessel image includes multiple frames of two-dimensional images, the smallest first width of the first region in the radial direction in the multiple frames of two-dimensional images and the The difference between the smallest second widths in the radial direction of a region determines the stenosis degree value, which can facilitate rapid quantification of the stenosis degree value of the blood vessel.
  • the three-dimensional blood vessel image includes multiple frames of two-dimensional images, and the two-dimensional images are detected to obtain an area detection result of the two-dimensional images, including: respectively using each frame of the two-dimensional image as an image to be measured, and comparing the image to be measured and the image to be measured with the image to be measured.
  • the two-dimensional images within a preset number of frames are detected at the image interval, and the area detection result of the image to be detected is obtained.
  • the three-dimensional blood vessel image includes multiple frames of two-dimensional images, so that each frame of the two-dimensional image is used as the image to be measured, and the image to be measured and the two-dimensional images within a preset number of frames spaced from the image to be measured are detected to obtain the image to be measured. Therefore, the continuous information between frames in the two-dimensional image can be used for detection, which is beneficial to improve the accuracy of image detection.
  • the area detection result further includes the first area corresponding to the blood vessel; after obtaining the area detection result of the image to be tested, the method further includes: according to the sequence of the multiple frames of two-dimensional images in the three-dimensional blood vessel image, the detected multiple splicing frames of two-dimensional images to obtain a three-dimensional detection image corresponding to the three-dimensional blood vessel image; wherein, the three-dimensional detection image includes a second area obtained by splicing the first areas of multiple frames of two-dimensional images, and the first area of the multiple frames of two-dimensional images. The second sub-region obtained by splicing the sub-regions; using the second region and the second sub-region to determine the position of the plaque in the blood vessel of the object to be measured.
  • the area detection result also includes the first area corresponding to the blood vessel, so that the detected two-dimensional images of the multiple frames are spliced according to the sequence of the multiple frames of two-dimensional images in the three-dimensional blood vessel image to obtain the corresponding three-dimensional blood vessel image.
  • a three-dimensional detection image, and the three-dimensional detection image includes a second region obtained by splicing the first regions of multiple frames of two-dimensional images, and a second sub-region obtained by splicing the first sub-regions of multiple frames of two-dimensional images, and then using the second region and the second sub-area to determine the location of plaque in the blood vessel of the object to be measured, so the regional detection results of multiple frames of two-dimensional images can be fused to locate the plaque in the blood vessel of the object to be measured, which is beneficial to improve user experience.
  • the blood vessel image detection method further includes: using the volume of the second sub-region and the volume of the second region to determine the stenosis degree value of the blood vessel of the object to be measured.
  • the stenosis degree value of the blood vessel to be measured is determined by the volume of the second sub-region and the volume of the second region, which can facilitate rapid quantification of the stenosis degree value of the blood vessel.
  • the two-dimensional images within a preset number of frames apart from the image to be measured include at least one of the following: a two-dimensional image located before the image to be measured, and a two-dimensional image located after the image to be measured.
  • the two-dimensional images within the preset frame number range from the image to be measured to include at least one of the following: a two-dimensional image located before the image to be measured, and a two-dimensional image located after the image to be measured, it is possible to use the two-dimensional image to be measured.
  • the continuous information between frames before the measurement image, or the continuous information between the frames after, or the continuous information between the frames before and after the image is detected, which can help to improve the accuracy of the image detection.
  • the area detection result is obtained by using the detection model to detect the two-dimensional image; and/or, the two-dimensional image is detected to obtain the area detection result of the two-dimensional image, including: performing a first detection on the two-dimensional image, determining Whether there is a first area corresponding to the blood vessel in the two-dimensional image; if the first area corresponding to the blood vessel is detected, perform a second detection on the first area in the two-dimensional image to obtain the first area in the first area. sub area.
  • the area detection result is obtained by using the detection model to detect the two-dimensional image, which can help to improve the efficiency of image detection; and by performing the first detection on the two-dimensional image, it is determined whether there is a second blood vessel corresponding to the blood vessel in the two-dimensional image. a region, and when the first region corresponding to the blood vessel is detected, the second detection is performed on the first region of the two-dimensional image to obtain the first sub-region in the first region, and the detection of the two-dimensional image can be divided into There are two steps for blood vessel detection and plaque detection, which can help improve the robustness of image detection.
  • performing the first detection on the two-dimensional image to determine whether there is a first region corresponding to the blood vessel in the two-dimensional image includes: using the first detection sub-network of the detection model to perform the first detection on the two-dimensional image, and determining the two-dimensional image Whether there is a first region corresponding to the blood vessel in the two-dimensional image; performing a second detection on the first region in the two-dimensional image to obtain the first sub-region in the first region, including: using the second detection sub-network of the detection model to perform a second detection on the two-dimensional image. The second detection is performed on the first region in the image to obtain the first sub-region within the first region.
  • the first detection sub-network of the detection model to perform the first detection on the two-dimensional image, it is determined whether there is a first region corresponding to the blood vessel in the two-dimensional image, and the second detection sub-network of the detection model is used to detect the two-dimensional image.
  • the second detection is performed on the first region in the first region, and the first sub-region in the first region is obtained, which can help to improve the efficiency of image detection.
  • the method further includes: in the case where the first region corresponding to the blood vessel is not detected, for the next frame of the two-dimensional image The image performs the first detection.
  • the network structure of the detection model can be simplified; Performing the first detection on the next frame of the two-dimensional image can help improve the efficiency of image detection.
  • a second aspect of the present application provides a method for training a detection model, including: acquiring at least one frame of sample two-dimensional image, and the at least one frame of sample two-dimensional image is a direction perpendicular to the extending direction of the eye and the blood vessel from the sample three-dimensional blood vessel image Extracted, the sample two-dimensional image is marked with the first actual sub-region corresponding to the plaque in the blood vessel; the detection model is used to detect the sample two-dimensional image, and the sample two-dimensional image corresponding to the plaque in the blood vessel is obtained.
  • the first prediction sub-region; the network parameters of the detection model are adjusted by using the difference between the first actual sub-region and the first prediction sub-region.
  • the sample two-dimensional image is marked with the blood vessel in the The first actual sub-region corresponding to the plaque, and the detection model is used to detect the sample two-dimensional image to obtain the first predicted sub-region corresponding to the plaque in the blood vessel in the sample two-dimensional image, so as to use the first actual sub-region
  • the difference from the first prediction sub-region, adjusting the network parameters of the detection model can make the detection model detect the pixels belonging to the plaques in the blood vessels, so as to realize the detection at the pixel level, which can improve the detection accuracy of the plaques in the blood vessels, and can It is beneficial to provide data support for the subsequent quantification of the stenosis degree value of blood vessels.
  • the sample three-dimensional blood vessel image includes multiple frames of sample two-dimensional images; the sample two-dimensional image is detected by using a detection model to obtain a first prediction sub-region corresponding to the plaque in the blood vessel in the sample two-dimensional image, including:
  • the two-dimensional image of the frame sample is used as the sample image to be tested, and the detection model is used to detect the sample to be tested image and the sample two-dimensional image within a preset number of frames from the sample to be tested image to obtain the first prediction in the sample to be tested image. sub area.
  • At least one frame of the sample two-dimensional image is extracted from the sample three-dimensional blood vessel image, and the sample three-dimensional blood vessel image includes multiple frames of the sample two-dimensional image, so that each frame of the sample two-dimensional image is taken as the sample image to be tested, and the The detection model detects the sample to-be-tested image and the sample two-dimensional image within a preset number of frames apart from the sample to-be-tested image to obtain the first prediction sub-region in the sample to-be-tested image, and then can use the frame interval in the sample two-dimensional image to be tested. Detecting continuous information is beneficial to improve the accuracy of image detection.
  • using the detection model to detect the sample two-dimensional image to obtain the first prediction sub-region corresponding to the plaque in the blood vessel in the sample two-dimensional image including: using the first detection sub-network of the detection model to perform the sample two-dimensional image
  • the first detection is to determine whether there is a first prediction area corresponding to the blood vessel in the two-dimensional image of the sample; in the case where the first prediction area corresponding to the blood vessel is detected, the second detection sub-network of the detection model is used to detect the sample two-dimensional image.
  • the second detection is performed on the first prediction area in the first prediction area to obtain the first prediction sub-area in the first prediction area.
  • the first detection sub-network of the detection model to perform the first detection on the sample two-dimensional image, it is determined whether there is a first prediction area corresponding to the blood vessel in the sample two-dimensional image, and when the first prediction area corresponding to the blood vessel is detected
  • the second detection sub-network of the detection model is used to perform a second detection on the first prediction region in the two-dimensional sample image to obtain the first prediction sub-region in the first prediction region, which can convert the
  • the detection is divided into two steps: blood vessel detection and plaque detection, which can help improve the robustness of image detection.
  • a third aspect of the present application provides a blood vessel image detection device, comprising: an image acquisition module, an image extraction module and a region detection module, the image acquisition module is configured to acquire a three-dimensional blood vessel image of an object to be measured; the image extraction module is configured to At least one frame of two-dimensional image is extracted from the three-dimensional blood vessel image in the direction perpendicular to the extending direction of the blood vessel in the blood vessel image; the area detection module is configured to detect the two-dimensional image to obtain the area detection result of the two-dimensional image, wherein the area detection result A first sub-region corresponding to the plaque in the blood vessel is included.
  • a fourth aspect of the present application provides a training device for a detection model, including: a sample acquisition module, a region detection module, and a parameter adjustment module, the sample acquisition module is configured to acquire at least one frame of a two-dimensional sample image, and at least one frame of a two-dimensional sample image The image is extracted from the direction perpendicular to the extension direction of the eye and the blood vessel in the sample three-dimensional blood vessel image, and the sample two-dimensional image is marked with a first actual sub-region corresponding to the plaque in the blood vessel; the area detection module is configured to use detection.
  • the model detects the sample two-dimensional image, and obtains the first predicted sub-region corresponding to the plaque in the blood vessel in the sample two-dimensional image; the parameter adjustment module is configured to use the difference between the first actual sub-region and the first predicted sub-region to adjust Detect the network parameters of the model.
  • a fifth aspect of the present application provides an electronic device, including a memory and a processor coupled to each other, the processor is configured to execute program instructions stored in the memory, so as to implement the blood vessel image detection method in the first aspect above, or to implement the above The training method of the detection model in the second aspect.
  • a sixth aspect of the present application provides a computer-readable storage medium on which program instructions are stored, and when the program instructions are executed by a processor, implement the blood vessel image detection method in the first aspect above, or implement the detection method in the second aspect above The training method of the model.
  • the region detection result of the two-dimensional image is obtained, and the region detection result includes the first sub-region corresponding to the plaque in the blood vessel, so the pixels belonging to the plaque in the blood vessel in the three-dimensional blood vessel image can be detected, and the detection at the pixel level can be realized.
  • the detection accuracy of the plaque in the blood vessel can be improved, and the data support can be provided for the subsequent quantification of the stenosis degree value of the blood vessel.
  • FIG. 1 is a schematic flowchart of an embodiment of a blood vessel image detection method of the present application
  • FIG. 2 is a schematic diagram of an embodiment of detecting a two-dimensional image
  • FIG. 3 is a schematic diagram of an embodiment of a three-dimensional detection image
  • FIG. 4 is a schematic flowchart of another embodiment of the blood vessel image detection method of the present application.
  • FIG. 5 is a schematic flowchart of an embodiment of a training method for a detection model of the present application
  • FIG. 6 is a schematic structural diagram of an embodiment of a blood vessel image detection device of the present application.
  • FIG. 7 is a schematic structural diagram of an embodiment of a training device for a detection model of the present application.
  • FIG. 8 is a schematic structural diagram of an embodiment of an electronic device of the present application.
  • FIG. 9 is a schematic structural diagram of an embodiment of a computer-readable storage medium of the present application.
  • system and “network” are often used interchangeably herein.
  • the term “and/or” in this article is only an association relationship to describe the associated objects, indicating that there can be three kinds of relationships, for example, A and/or B, it can mean that A exists alone, A and B exist at the same time, and A and B exist independently B these three cases.
  • the character "/” in this document generally indicates that the related objects are an “or” relationship.
  • “multiple” herein means two or more than two.
  • Embodiments of the present disclosure provide a blood vessel image detection method, the execution subject of which may be a blood vessel image detection apparatus.
  • the blood vessel image detection method may be executed by a terminal device or a server or other electronic device, where the terminal device may be a user equipment ( User Equipment, UE), mobile devices, user terminals, terminals, cellular phones, cordless phones, personal digital assistants (Personal Digital Assistant, PDA), handheld devices, computing devices, in-vehicle devices, wearable devices, etc.
  • the blood vessel image detection method may be implemented by the processor calling computer-readable instructions stored in the memory.
  • FIG. 1 is a schematic flowchart of an embodiment of a blood vessel image detection method of the present application. Specifically, the following steps can be included:
  • Step S11 Acquire a three-dimensional blood vessel image of the object to be measured.
  • a three-dimensional blood vessel image of the object to be measured can be obtained by scanning the blood vessels in the preset part of the object to be measured.
  • the preset position can be set according to actual application requirements, and specifically may include but not limited to: cardiac coronary arteries, internal carotid arteries, external carotid arteries, vertebral arteries, etc., which are not limited herein. Since the blood vessels usually travel in a distorted way and are not on the same plane, the scanned medical images (such as computed tomography, etc.) can be straightened using Curved Planar Reformation (CPR) to obtain a straightened CPR image, and the It serves as a three-dimensional blood vessel image of the object to be measured.
  • CPR Curved Planar Reformation
  • initial detection can be performed on the scanned three-dimensional blood vessel image to obtain blood vessels in the three-dimensional blood vessel image.
  • neural networks eg, U-Net, etc.
  • traditional image segmentation methods eg, edge extraction, and morphological treatment methods such as expansion corrosion
  • details are not repeated here. Therefore, the center line of the blood vessel can be determined, and along the center line of the blood vessel, the cross-sectional images of the blood vessel can be extracted, and the cross-sectional images belonging to the same blood vessel can be stacked in sequence, so that the three-dimensional blood vessel image of the object to be measured in the embodiment of the present disclosure can be obtained. .
  • Step S12 Extract at least one frame of two-dimensional image from the three-dimensional blood vessel image along a direction perpendicular to the extending direction of the blood vessel in the three-dimensional blood vessel image.
  • the extending direction of the blood vessel refers to the tangential direction of the center line of the blood vessel.
  • two-dimensional images can be extracted at various positions of the blood vessel along the normal plane of its centerline.
  • an image to be processed can be extracted at each position of the blood vessel along the normal plane of its center line, and the to-be-processed image can be extracted.
  • a two-dimensional image of a preset size eg, 64*64 is obtained.
  • Step S13 Detect the two-dimensional image to obtain a region detection result of the two-dimensional image, wherein the region detection result includes a first sub-region corresponding to the plaque in the blood vessel.
  • each frame of the two-dimensional image may be independently detected to obtain a region detection result of the two-dimensional image.
  • edge detection can be performed on a two-dimensional image to obtain a mask image containing edge information (eg, blood vessel edge information, plaque edge information, etc.) By removing abnormal pixels such as outliers in the mask image, the region detection results of the two-dimensional image can be obtained.
  • edge information eg, blood vessel edge information, plaque edge information, etc.
  • the three-dimensional blood vessel image may be extracted to obtain multiple frames of two-dimensional images, for example, 50 frames of two-dimensional images, 100 frames of two-dimensional images, 150 frames of two-dimensional images, etc., which are not limited herein. Then, each frame of the two-dimensional image can also be used as the image to be tested, and the image to be tested and the two-dimensional images within a preset number of frames apart from the image to be tested can be detected to obtain the area detection result of the image to be tested. In the above manner, continuous information between frames in a two-dimensional image can be used for detection, which is beneficial to improve the accuracy of image detection.
  • the preset number of frames may be set according to actual application requirements, for example: 1, 2, 3, 4, etc., which are not limited herein.
  • the preset number of frames is 0, that is, the 2D image within the preset number of frames separated from the image to be tested is the image to be tested itself.
  • the two-dimensional image within a preset number of frames from the image to be measured may include at least one of the following: a two-dimensional image located before the image to be measured, and a two-dimensional image located after the image to be measured. That is, the two-dimensional image within a preset number of frames from the image to be tested may be a two-dimensional image located before the image to be tested and within a preset number of frames from the image to be tested, for example, a two-dimensional image located within 3 frames before the image to be tested.
  • the two-dimensional image within a preset number of frames from the image to be tested may also be a two-dimensional image located after the image to be tested and within a preset number of frames from the image to be tested, for example, after the image to be tested Two-dimensional images within 3 frames; or, two-dimensional images within a preset number of frames spaced from the image to be tested may also include: a two-dimensional image located before the image to be tested and within a preset number of frames spaced from the image to be tested and a two-dimensional image located within a preset number of frames from the image to be tested.
  • the two-dimensional image after the image to be tested and within a preset number of frames from the image to be tested for example, the two-dimensional image located within 3 frames before the image to be tested and the two-dimensional image located within 3 frames after the image to be tested.
  • FIG. 2 is a schematic diagram of an embodiment of detecting a two-dimensional image.
  • a detection model can be pre-trained, so that the two-dimensional image can be detected by using the detection model.
  • the detection is performed to obtain the region detection result of the two-dimensional image.
  • the detection model can specifically use U-Net, Fully Convolutional Network (FCN), etc., which is not limited here.
  • FCN Fully Convolutional Network
  • the convolution kernel used for extracting image features in the detection model can be specifically set as a two-dimensional convolution kernel, then each frame of two-dimensional image can be detected separately to obtain the region detection result of the two-dimensional image, or each frame of two-dimensional image can be detected separately.
  • the frame two-dimensional image is used as the image to be tested, and the image to be tested and the two-dimensional image within a preset number of frames spaced from the image to be tested are detected to obtain the area detection result of the image to be tested.
  • the convolution kernel used for extracting image features in the detection model can be specifically set as a three-dimensional convolution kernel, then each frame of two-dimensional image can be used as the image to be measured, and the image to be measured and the image to be measured are separated by preset frames.
  • the two-dimensional images within the number of 2D images are detected to obtain the area detection results of the image to be tested and the two-dimensional images within a preset number of frames apart from the image to be tested.
  • it can be set according to actual application requirements, which is not limited here.
  • the detection model may be used to detect each two-dimensional image respectively to obtain a region detection result of the two-dimensional image.
  • each frame of the two-dimensional image can also be used as the image to be measured, and the detection model can be used to detect the image to be measured and the two-dimensional images within a preset number of frames from the image to be measured, and obtain the image to be measured. Therefore, the continuous information between frames in the two-dimensional image can be used for detection, which is beneficial to improve the accuracy of image detection.
  • the detection model can be used to detect the image to be measured and the two-dimensional images within a preset number of frames from the image to be measured, and obtain the image to be measured. Therefore, the continuous information between frames in the two-dimensional image can be used for detection, which is beneficial to improve the accuracy of image detection.
  • the area detection result may further include a first area corresponding to the blood vessel.
  • the obliquely filled part is the first area corresponding to the blood vessel
  • the white filled part in the first area is the first area corresponding to the blood vessel.
  • the first sub-region corresponding to the block, the stenosis degree value of the blood vessel to be measured can be obtained based on the first region and the first sub-region in at least one frame of two-dimensional image. In the above manner, the stenosis degree of the blood vessel can be quickly counted and quantified according to the detected first sub-region and the first region, which is beneficial to improve user experience.
  • the stenosis degree value may be determined by using the area of the first sub-region and the area of the first region. For example, the ratio between the detected area of the first sub-region and the area of the first region can be used as the stenosis degree value.
  • the larger the stenosis degree value the higher the degree of blockage of the blood vessel by the plaque.
  • the higher the degree of vascular stenosis on the contrary, the smaller the value of stenosis degree, the lower the degree of occlusion of the blood vessel by plaque, and the lower the degree of vascular stenosis.
  • the degree of narrowing can also be determined by using the smallest first width D of the first region in the radial direction and the smallest second width d of the first sub-region in the radial direction of the first region value. It should be noted that, if the blood vessel center line passes through a pixel point in the two-dimensional image (as shown in the circle filled with black in Figure 2), the direction passing through the pixel point in the two-dimensional image is the radial direction. Please refer to FIG. 2 , the second width of the first sub-region of the patch in a radial direction of the first region (indicated by the double-headed arrow in FIG.
  • the first width is D1
  • the second width of the first sub-region of the patch in the other radial direction of the first region is d2
  • the first region is in this radial direction
  • the first width is D2
  • so on, in different radial directions, the second width of the first region of the patch and the first width of the first region can be obtained respectively, and the smallest second width d and The smallest first width D, the ratio of the second width d to the first width D can be used as the stenosis degree value in the radial direction.
  • the larger the stenosis degree value the more the blockage of the blood vessel by plaque.
  • the higher the degree the higher the degree of stenosis of the blood vessel.
  • the smaller the value of the degree of stenosis the lower the degree of occlusion of the blood vessel by the plaque, and the lower the degree of vascular stenosis.
  • other methods can also be used to calculate the stenosis degree value.
  • the NASCET North American Symptomatic Carotid Endarterectomy, North American Symptomatic Carotid Endarterectomy Test
  • ECST MRC European Carotid Surgery Trial, European Carotid Surgery Trial
  • Other scenarios can be set according to the actual application, so I will not give examples here.
  • the difference between the area of the first region and the area of the first sub-region in the multiple frames of two-dimensional images can also be used to determine the stenosis degree value.
  • the difference S-s between the area of the first region and the area S of the first sub-region in each frame of the two-dimensional image and the area s of the first sub-region can be obtained as the effective area of the blood vessel in the two-dimensional image.
  • S' it can be denoted as
  • the effective area S' of the blood vessel in the multi-frame two-dimensional image should change steadily with the extension of the blood vessel.
  • the effective area S' of blood vessels in multi-frame two-dimensional images should change with the extension of blood vessels.
  • a preset method can be used to fit the effective area S' of the blood vessels in the multi-frame two-dimensional images to obtain the theoretical value of the effective area S', and the actual measured value of the effective area S' of the blood vessel can be calculated , as the actual value of the effective area S' of the blood vessel, so that the percentage of the actual value deviating from the theoretical value can be taken as the value of the stenosis degree of the blood vessel, the larger the calculated stenosis degree value, the more stenotic the blood vessel; Taking the ratio of the actual value of the effective area S' to the theoretical value of the effective area S' as the stenosis degree value of the blood vessel, the smaller the calculated stenosis degree value, the more stenotic the blood vessel is.
  • the smallest first width of the first region in the radial direction and the first sub-region in the multiple frames of two-dimensional images may also be used.
  • the difference between the smallest second widths in the radial direction of the first region determines the stenosis degree value.
  • the smallest second width d of the first sub-region in the radial direction of the first region and the smallest first width D of the first region in each frame of two-dimensional image can be obtained, and the first width D of the blood vessel can be compared with the first width D of the blood vessel.
  • the difference D-d of the second width d of the plaque as the effective width of the blood vessel in the two-dimensional image, may be denoted as D' for the convenience of description.
  • D' the effective width of the blood vessel in the multi-frame two-dimensional image should change steadily (eg, linearly increase, or linearly decrease) with the extension of the blood vessel, for example,
  • the effective width D' of the blood vessel in the first frame of two-dimensional image is 2 mm
  • the effective width D' of the blood vessel in the second frame of two-dimensional image is 1.9 mm
  • the effective width D' of the blood vessel in the third frame of two-dimensional image is 1.8 mm
  • the effective width D' of the blood vessel in the first frame of two-dimensional image is 2mm
  • the effective width D' of the blood vessel in the second frame of two-dimensional image is 1.9mm
  • the effective width D' of the blood vessel in the third frame of two-dimensional image is 1.6mm
  • the effective width D' of the blood vessel in the fourth frame of two-dimensional image is 1.3mm
  • the effective width D' of the blood vessel in the fifth frame of the two-dimensional image is 0.9 mm. It can be seen that the effective width D' presents a sudden change with the extension of the blood vessel.
  • a preset method can be used to perform linear fitting on the effective width D' of the blood vessels in the multi-frame two-dimensional images, to obtain the value of the effective width D' that changes linearly with the extension of the blood vessel, as the effective width D' of the blood vessel.
  • the theoretical value of the blood vessel, and the actual measured value of the effective width D' of the blood vessel is taken as the actual value of the effective width D' of the blood vessel, so that the percentage of the actual value deviating from the theoretical value can be regarded as the value of the stenosis degree of the blood vessel, then calculate The greater the obtained stenosis degree value, the more stenotic the blood vessel is; alternatively, the ratio of the actual value of the effective width D' to the theoretical value of the effective width D' can be directly used as the stenosis degree value, and the calculated stenosis degree value is higher. Smaller, indicating narrower blood vessels.
  • the detection report may also be generated by using the stenosis degree value.
  • the detection report may be set in a preset format, and the detected stenosis degree value may be added to the field corresponding to the stenosis degree value in the detection report.
  • the images marked with the detection results of the regions can be attached to the detection report as attachments.
  • the area detection result may further include a first area corresponding to the blood vessel
  • the detected multiple frames of two-dimensional images may also be processed according to the sequence of the multiple frames of two-dimensional images in the three-dimensional blood vessel image. splicing to obtain a three-dimensional detection image corresponding to the three-dimensional blood vessel image, and the three-dimensional detection image includes a second region obtained by splicing the first regions of multiple frames of two-dimensional images and a second region obtained by splicing the first sub-regions of multiple frames of two-dimensional images. sub-region, so that the second region and the second sub-region can be used to determine the position of the plaque in the blood vessel of the object to be tested.
  • the regional detection results of multiple frames of two-dimensional images can be fused to locate plaques in the blood vessels of the object to be measured, which is beneficial to improve user experience.
  • the stenosis degree value may be determined by further utilizing the volume of the second sub-region and the volume of the second region.
  • the volume of the second region may be the number of pixels included in the second region, and similarly, the volume of the second subregion may be the number of pixels included in the second subregion.
  • FIG. 3 is a schematic diagram of an embodiment of a three-dimensional detection image.
  • the cuboid area represents a three-dimensional detection image formed by splicing multiple frames of two-dimensional images
  • the cylindrical area in the cuboid area represents the first area formed by splicing
  • the second area corresponding to the blood vessel, the small dot filled area represents the second sub-area corresponding to the plaque formed by the splicing of the first sub-area. Positioning to improve user experience.
  • the area detection result may further include the specific type of the detected plaque, such as calcified plaque, mixed plaque, non-calcified plaque, etc., which is not limited herein.
  • the region detection result of the two-dimensional image is obtained, and the region detection result includes the first sub-region corresponding to the plaque in the blood vessel, so the pixels belonging to the plaque in the blood vessel in the three-dimensional blood vessel image can be detected, and the detection at the pixel level can be realized.
  • the detection accuracy of the plaque in the blood vessel can be improved, and the data support can be provided for the subsequent quantification of the stenosis degree value of the blood vessel.
  • FIG. 4 is a schematic flowchart of another embodiment of the blood vessel image detection method of the present application. Specifically, the following steps may be included:
  • Step S41 acquiring a three-dimensional blood vessel image of the object to be measured.
  • Step S42 Extract at least one frame of two-dimensional image from the three-dimensional blood vessel image along a direction perpendicular to the extending direction of the blood vessel in the three-dimensional blood vessel image.
  • Step S43 Perform a first detection on the two-dimensional image to determine whether there is a first region corresponding to the blood vessel in the two-dimensional image.
  • edge detection can be performed on a two-dimensional image to obtain a mask image containing edge information, and morphological processing such as dilation corrosion can be performed on the mask image to filter out outliers in the mask image. Therefore, the edge line in the two-dimensional image can be obtained, and the area enclosed by the edge line can be used as the closed area corresponding to the edge line, and then the largest edge line of the closed area can be used as the edge line of the blood vessel, and The closed area is taken as the first area corresponding to the blood vessel.
  • the first detection sub-network of the detection model can also be used to perform the first detection on the two-dimensional image to determine whether there is a first region corresponding to the blood vessel in the two-dimensional image, so that the detection model can be used for detection, Improve detection efficiency.
  • the detection model can also be used to perform the first detection on the two-dimensional image to determine whether there is a first region corresponding to the blood vessel in the two-dimensional image, so that the detection model can be used for detection, Improve detection efficiency.
  • each frame of the two-dimensional image can also be used as the image to be tested, and the image to be tested and the two-dimensional image within a preset number of frames from the image to be tested are used for detection to determine whether the two-dimensional image is There is a first region corresponding to a blood vessel.
  • the first detection sub-network of the detection model can be used to perform the first detection on the image to be tested and the two-dimensional image within a preset number of frames from the image to be tested, to determine whether there is a first region corresponding to the blood vessel in the two-dimensional image.
  • the convolution kernel used for extracting image features in the detection model reference may be made to the relevant descriptions in the foregoing disclosed embodiments, and details are not repeated here.
  • Step S44 in the case where the first region corresponding to the blood vessel is detected, perform a second detection on the first region in the two-dimensional image to obtain a first sub-region in the first region.
  • the second detection is performed on the first region of the two-dimensional image to obtain the first sub-region in the first region, that is, the first sub-region corresponding to the plaque. first subregion.
  • edge detection may be used to perform the second detection on the first region, so as to obtain the first sub-region corresponding to the plaque in the first region.
  • the related description of a region will not be repeated here.
  • the second detection sub-network of the detection model may also be used to perform second detection on the first region in the two-dimensional image to obtain the first sub-region within the first region.
  • the first detection sub-network and the second detection sub-network may have the same network structure, so that the network structure of the detection model can be simplified.
  • the first detection sub-network and the second detection sub-network can be both U-Nets, that is, the detection model is composed of two U-Nets connected in series, and other scenarios can be set according to actual application needs, which will not be described here. limited.
  • the first detection may be performed on the next frame of the two-dimensional image.
  • the first detection can be directly performed on the next frame of two-dimensional image to detect whether there is a blood vessel in the next frame of two-dimensional image.
  • the first area corresponding to the blood vessel can improve the detection efficiency.
  • the first detection on the two-dimensional image it is determined whether there is a first area corresponding to the blood vessel in the two-dimensional image, and when the first area corresponding to the blood vessel is detected, the two-dimensional image is detected.
  • the second detection is performed on the first region of the first region to obtain the first sub-region in the first region, and the detection of the two-dimensional image can be divided into two steps: blood vessel detection and plaque detection, which can help improve the robustness of image detection.
  • FIG. 5 is a schematic flowchart of an embodiment of a training method for an image detection model of the present application. Specifically, the following steps may be included;
  • Step S51 Acquire at least one frame of the sample two-dimensional image, at least one frame of the sample two-dimensional image is extracted from the sample three-dimensional blood vessel image along the direction perpendicular to the extending direction of the blood vessel, and the sample two-dimensional image is marked with the blood vessel.
  • the patch corresponds to the first actual subregion.
  • the sample two-dimensional image may be extracted from the sample three-dimensional blood vessel image along a direction perpendicular to the extending direction of the blood vessel.
  • the sample three-dimensional blood vessel image of the sample For the specific acquisition method of the three-dimensional blood vessel image of the sample, reference may be made to the relevant descriptions in the foregoing disclosed embodiments, which will not be repeated here.
  • the specific extraction method reference may be made to the method of extracting at least one frame of two-dimensional image from the three-dimensional blood vessel image in the foregoing disclosed embodiments, and details are not described herein again.
  • the two-dimensional image of the sample may also be marked with specific types of plaques, such as calcified plaques, mixed plaques, non-calcified plaques, etc., which are not limited herein.
  • the two-dimensional image of the sample may also be marked with a first actual region corresponding to the blood vessel.
  • Step S52 Use the detection model to detect the sample two-dimensional image to obtain a first predicted sub-region corresponding to the plaque in the blood vessel in the sample two-dimensional image.
  • the detection model may be specifically set according to actual application requirements, and for details, reference may be made to the relevant descriptions in the foregoing disclosed embodiments, which will not be repeated here.
  • the detection model may also detect the sample two-dimensional image to obtain the first predicted region corresponding to the blood vessel in the sample two-dimensional image.
  • the detection model may specifically include a first detection sub-network and a second detection sub-network, so that the first detection sub-network of the detection model can be used to perform the first detection on the sample two-dimensional image, and determine the sample two-dimensional image. Whether there is a first prediction area corresponding to the blood vessel in the image, and when the first prediction area corresponding to the blood vessel is detected, use the second detection sub-network of the detection model to perform the first prediction area in the sample two-dimensional image. In the second detection, the first prediction sub-region in the first prediction region is obtained, which can help to improve the detection efficiency.
  • the first detection sub-network and the second detection sub-network may have the same network structure.
  • the first detection sub-network and the second detection sub-network are both U-Nets, that is, the detection model is composed of a series of U-Nets. composition, other scenarios can also be set according to actual application needs, which is not limited here.
  • the second detection sub-network of the detection model can also be used to perform the above-mentioned first detection on the two-dimensional image of the next frame of samples. For details, please refer to the aforementioned disclosed embodiments The relevant descriptions in , will not be repeated here.
  • the above-mentioned sample three-dimensional blood vessel image includes multiple frames of sample two-dimensional images, then each frame of sample two-dimensional image can be used as the sample image to be tested, and the sample to be tested image and the sample to be tested image can be compared using the detection model.
  • the two-dimensional images of the samples within a preset number of frames are detected to obtain the first prediction sub-region in the sample to-be-measured image, so that continuous information between frames can be combined for detection, which is beneficial to improve the detection accuracy.
  • the first detection sub-network of the detection model can be used to detect the sample to be tested image and the sample two-dimensional image within a preset number of frames from the sample to be tested image.
  • There is a first prediction area corresponding to the blood vessel and when the first prediction area corresponding to the blood vessel is detected, the second detection sub-network of the detection model is used to perform a second detection on the first prediction area in the sample two-dimensional image , to obtain the first prediction sub-region in the first prediction region.
  • Step S53 Adjust the network parameters of the detection model by using the difference between the first actual sub-region and the first predicted sub-region.
  • a preset loss function (such as a cross-entropy loss function, a dice loss function) can be used to process the first actual sub-region and the first predicted sub-region to obtain the first loss value of the detection model, so as to utilize the first loss The value adjusts the network parameters of the detection model.
  • the first actual region corresponding to the blood vessel is also marked in the sample two-dimensional image, and the detection model also detects the first predicted region corresponding to the blood vessel, and a preset loss function (such as a cross entropy loss function) can also be used.
  • the detection model also detects the predicted type of the patch, and a preset loss function (such as a cross-entropy loss function) can be used to process the actual type and predicted type, and get The third loss value of the detection model can be used to adjust the network parameters of the detection model by using the first loss value, the second loss value and the third loss value.
  • a preset loss function such as a cross-entropy loss function
  • the preset loss function may include, but is not limited to, a cross entropy loss function (cross entropy loss), dice loss, etc., which are not limited herein.
  • Stochastic Gradient Descent SGD
  • Batch Gradient Descent BGD
  • Mini-Batch Gradient Descent MBGD
  • Batch gradient descent refers to using all samples to update parameters in each iteration; stochastic gradient descent refers to using one sample for each iteration. Parameter update; Mini-batch gradient descent refers to using a batch of samples to update parameters in each iteration, which will not be repeated here.
  • a training end condition may also be set, and when the training end condition is satisfied, the training of the detection model may be ended.
  • the training termination conditions may include: the loss value is less than a preset loss threshold, and the loss value is no longer reduced; the current training times reaches the preset times threshold (for example, 500 times, 1000 times, etc.), which is not limited here .
  • a second sample is obtained.
  • the first prediction sub-region corresponding to the plaque in the blood vessel in the 3D image so that the difference between the first actual sub-region and the first predicted sub-region can be used to adjust the network parameters of the detection model, so that the detection model can detect the plaque belonging to the blood vessel.
  • the pixel points of the block are used to achieve pixel-level detection, which can improve the detection accuracy of plaques in blood vessels, and can help to provide data support for subsequent quantification of blood vessel stenosis degree values.
  • FIG. 6 is a schematic structural diagram of an embodiment of a blood vessel image detection apparatus 60 of the present application.
  • the blood vessel image detection device 60 includes: an image acquisition module 61, an image extraction module 62 and a region detection module 63, the image acquisition module 61 is configured to acquire a three-dimensional blood vessel image of the object to be measured; the image extraction module 62 is configured to be along the blood vessels in the three-dimensional blood vessel image. At least one frame of two-dimensional image is extracted from the three-dimensional blood vessel image in a direction perpendicular to the extension direction of The plaques in the corresponding first sub-regions.
  • the detection accuracy of plaque in the blood vessel can be improved, and data support can be provided for the subsequent quantification of the stenosis degree value of the blood vessel.
  • the region detection result further includes a first region corresponding to the blood vessel
  • the blood vessel image detection device 60 further includes a stenosis calculation module configured to be based on the first region and the first sub-region in the at least one frame of two-dimensional image , to obtain the stenosis degree value of the blood vessel to be measured.
  • the area detection result also includes a first area corresponding to the blood vessel, so that the stenosis degree value of the blood vessel to be measured can be obtained based on the first area and the first sub-area in at least one frame of two-dimensional image, thereby enabling Improve user experience.
  • the stenosis calculation module is configured to use the area of the first subregion and the area of the first region to determine the stenosis degree value; or, use the smallest first width and the first width of the first region in the radial direction the minimum second width of the sub-region in the radial direction of the first region, to determine the stenosis degree value; or, in the case where the three-dimensional blood vessel image includes multiple frames of two-dimensional images, use the area of the first region in the multiple frames of two-dimensional images
  • the stenosis degree value is determined by the difference with the area of the first sub-region; or, in the case where the three-dimensional blood vessel image includes multiple frames of two-dimensional images, the first area in the multiple frames of two-dimensional images is the smallest in the radial direction.
  • a difference between the width and the smallest second width of the first sub-region in the radial direction of the first region determines the stenosis degree value.
  • the stenosis degree value is determined by using the area of the first sub-region and the first width of the first region, or by using the maximum second width and the first width of the first region, to determine the stenosis degree value, or in the case where the three-dimensional blood vessel image includes multiple frames of two-dimensional images, use the difference between the area of the first region and the area of the first sub-region in the multiple frames of two-dimensional images , determine the stenosis degree value, or when the three-dimensional blood vessel image includes multiple frames of two-dimensional images, use the minimum first width of the first region in the radial direction in the multiple frames of two-dimensional images and the first sub-region in the first region The difference between the smallest second widths in the radial direction of , determines the stenosis degree value, which can help to quickly quantify the stenosis degree value of the blood vessel.
  • the three-dimensional blood vessel image includes multiple frames of two-dimensional images
  • the region detection module 63 is specifically configured to take each frame of the two-dimensional image as the image to be measured, and to separate the image to be measured and the image to be measured by a preset number of frames The two-dimensional image inside is detected, and the area detection result of the image to be tested is obtained.
  • the three-dimensional blood vessel image includes multiple frames of two-dimensional images, so that each frame of the two-dimensional image is used as the image to be measured, and the image to be measured and the two-dimensional image within a preset number of frames spaced from the image to be measured are detected. , to obtain the region detection result of the image to be tested, so the continuous information between frames in the two-dimensional image can be used for detection, which is beneficial to improve the accuracy of image detection.
  • the region detection result further includes a first region corresponding to the blood vessel
  • the blood vessel image detection device 60 further includes an image splicing module configured to, according to the sequence of the multiple frames of two-dimensional images in the three-dimensional blood vessel image, detect The obtained multiple frames of two-dimensional images are spliced to obtain a three-dimensional detection image corresponding to the three-dimensional blood vessel image; wherein, the three-dimensional detection image includes a second area obtained by splicing the first areas of the multiple frames of two-dimensional images, and multiple frames of two-dimensional images.
  • the second sub-region obtained by splicing the first sub-regions of the blood vessel image detection device 60 further includes a plaque locating module configured to use the second region and the second sub-region to determine the position of the plaque in the blood vessel of the object to be measured.
  • the area detection result also includes the first area corresponding to the blood vessel, so that the detected two-dimensional images are spliced according to the sequence of the two-dimensional images of the multiple frames in the three-dimensional blood vessel image to obtain the same three-dimensional image.
  • a three-dimensional detection image corresponding to the blood vessel image includes a second region obtained by splicing the first regions of multiple frames of two-dimensional images, and a second sub-region obtained by splicing the first sub-regions of the multiple frames of two-dimensional images, and then Using the second area and the second sub-area to determine the position of the plaque in the blood vessel of the object to be measured, it is possible to fuse the regional detection results of multiple frames of two-dimensional images to locate the plaque in the blood vessel of the object to be measured, which is conducive to improving the user experience. experience.
  • the stenosis calculation module is configured to use the volume of the second sub-region and the volume of the second region to determine the stenosis degree value of the blood vessel to be measured.
  • the stenosis degree value of the blood vessel to be measured is determined by the volume of the second sub-area and the volume of the second area, which can facilitate rapid quantification of the stenosis degree value of the blood vessel.
  • the two-dimensional images within a preset number of frames from the image to be measured include at least one of the following: a two-dimensional image located before the image to be measured, and a two-dimensional image located after the image to be measured.
  • image detection can be performed by using the inter-frame continuous information before the image to be tested, or the subsequent inter-frame continuous information, or the previous and subsequent inter-frame continuous information, which can help improve the accuracy of image detection.
  • the area detection result is obtained by using the detection model to detect the two-dimensional image; and/or, the area detection module 63 includes a first detection sub-module, configured to perform a first detection on the two-dimensional image, determine Whether there is a first region corresponding to the blood vessel in the two-dimensional image, the region detection module 63 includes a second detection sub-module, configured to detect the first region corresponding to the blood vessel in the case of detecting the first region in the two-dimensional image. A second detection is performed to obtain a first sub-region within the first region.
  • the area detection result is obtained by using the detection model to detect the two-dimensional image, which can help to improve the efficiency of image detection; and by performing the first detection on the two-dimensional image, it is determined whether there is a The first area corresponding to the blood vessel, and when the first area corresponding to the blood vessel is detected, the second detection is performed on the first area of the two-dimensional image to obtain the first sub-area in the first area, and the two-dimensional image can be converted into the first sub-area.
  • Image detection is divided into two steps: blood vessel detection and plaque detection, which can help improve the robustness of image detection.
  • the first detection sub-module is specifically configured to perform a first detection on the two-dimensional image by using the first detection sub-network of the detection model to determine whether there is a first region corresponding to a blood vessel in the two-dimensional image
  • the second The detection sub-module is specifically configured to use the second detection sub-network of the detection model to perform a second detection on the first region in the two-dimensional image to obtain the first sub-region in the first region.
  • the first detection sub-network of the detection model to perform the first detection on the two-dimensional image, it is determined whether there is a first region corresponding to the blood vessel in the two-dimensional image, and the second detection sub-network of the detection model is used.
  • the second detection is performed on the first region in the two-dimensional image to obtain the first sub-region in the first region, which can help improve the efficiency of image detection.
  • the first detection sub-network and the second detection sub-network have the same network structure, and/or the first detection sub-module is further configured to not detect the first region corresponding to the blood vessel , and perform the first detection on the next frame of the two-dimensional image.
  • the network structure of the detection model can be simplified; In this case, directly performing the first detection on the next frame of the two-dimensional image can help improve the efficiency of image detection.
  • FIG. 7 is a schematic structural diagram of an embodiment of a training device 70 for detecting a model of the present application.
  • the training device 70 for the detection model includes: a sample acquisition module 71, a region detection module 72 and a parameter adjustment module 73.
  • the sample acquisition module 71 is configured to acquire at least one frame of sample two-dimensional image, and at least one frame of the sample two-dimensional image is obtained from the sample three-dimensional blood vessel.
  • the image is extracted along the direction perpendicular to the extension direction of the blood vessel, and the sample two-dimensional image is marked with the first actual sub-region corresponding to the plaque in the blood vessel;
  • the area detection module 72 is configured to use the detection model to analyze the sample two-dimensional
  • the image is detected to obtain the first predicted sub-region corresponding to the plaque in the blood vessel in the two-dimensional image of the sample;
  • the parameter adjustment module 73 is configured to use the difference between the first actual sub-region and the first predicted sub-region to adjust the network of the detection model parameter.
  • At least one frame of the sample two-dimensional image is extracted from the sample three-dimensional blood vessel image along the direction perpendicular to the extending direction of the blood vessel, and the sample two-dimensional image is marked with the blood vessel.
  • the first actual sub-region corresponding to the plaque in the blood vessel, and the detection model is used to detect the sample two-dimensional image, and the first predicted sub-region corresponding to the plaque in the blood vessel in the sample two-dimensional image is obtained, so as to use the first actual sub-region.
  • the difference between the region and the first prediction sub-region can be adjusted by adjusting the network parameters of the detection model, so that the detection model can detect the pixels belonging to the plaques in the blood vessels, so as to achieve pixel-level detection, improve the detection accuracy of plaques in the blood vessels, and can It is beneficial to provide data support for the subsequent quantification of the stenosis degree value of blood vessels.
  • the sample three-dimensional blood vessel image includes multiple frames of sample two-dimensional images
  • the region detection module 72 is specifically configured to use each frame of the sample two-dimensional image as the sample image to be tested, and use the detection model to compare the sample to be tested image and the sample to be tested image. Detecting the sample two-dimensional image within a preset number of frames from the sample to-be-measured image to obtain a first prediction sub-region in the sample to-be-measured image.
  • the sample three-dimensional blood vessel image includes multiple frames of sample two-dimensional images, so that each frame of the sample two-dimensional image is taken as the sample image to be tested, and the sample to be tested image and the sample to be tested image are pre-measured by the detection model.
  • the sample two-dimensional images within the number of frames are detected to obtain the first prediction sub-region in the sample to-be-tested image, and the continuous information between frames in the sample two-dimensional image can be used for detection, which is beneficial to improve the accuracy of image detection.
  • the region detection module 72 includes a first detection sub-module configured to perform a first detection on the sample two-dimensional image by using the first detection sub-network of the detection model to determine whether there is a blood vessel in the sample two-dimensional image.
  • the first prediction area of The second detection is performed on the first prediction area to obtain a first prediction sub-area in the first prediction area.
  • the first detection sub-network of the detection model to perform the first detection on the sample two-dimensional image, it is determined whether there is a first predicted area corresponding to the blood vessel in the sample two-dimensional image, and when the blood vessel is detected.
  • the first prediction region of The detection of two-dimensional images is divided into two steps: blood vessel detection and plaque detection, which can help improve the robustness of image detection.
  • FIG. 8 is a schematic structural diagram of an embodiment of an electronic device 80 of the present application.
  • the electronic device 80 includes a memory 81 and a processor 82 that are coupled to each other, and the processor 82 is configured to execute program instructions stored in the memory 81 to implement the steps of any of the foregoing embodiments of the blood vessel image detection method, or to implement any of the foregoing detection models. The steps in the training method embodiment.
  • the electronic device 80 may include, but is not limited to, a microcomputer and a server.
  • the electronic device 80 may also include mobile devices such as a notebook computer and a tablet computer, which are not limited herein.
  • the processor 82 is configured to control itself and the memory 81 to implement the steps of any of the foregoing embodiments of the blood vessel image detection method, or to implement the steps of any of the foregoing embodiments of the detection model training method.
  • the processor 82 may also be referred to as a CPU (Central Processing Unit, central processing unit).
  • the processor 82 may be an integrated circuit chip with signal processing capability.
  • the processor 82 may also be a general-purpose processor, a digital signal processor (Digital Signal Processor, DSP), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array (Field-Programmable Gate Array, FPGA) or other Programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
  • a general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
  • the processor 82 may be jointly implemented by an integrated circuit chip.
  • the above solution can improve the detection accuracy of plaques in blood vessels.
  • FIG. 9 is a schematic structural diagram of an embodiment of a computer-readable storage medium 90 of the present application.
  • the computer-readable storage medium 90 stores program instructions 901 that can be executed by the processor, and the program instructions 901 are used to implement the steps of any of the above-mentioned embodiments of the blood vessel image detection method, or to implement any of the above-mentioned embodiments of the training method of the detection model. step.
  • the above solution can improve the detection accuracy of plaques in blood vessels.
  • the functions or modules included in the apparatuses provided in the embodiments of the present disclosure may be used to execute the methods described in the above method embodiments.
  • the disclosed method and apparatus may be implemented in other manners.
  • the device implementations described above are only illustrative.
  • the division of modules or units is only a logical function division. In actual implementation, there may be other divisions.
  • units or components may be combined or integrated. to another system, or some features can be ignored, or not implemented.
  • the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of devices or units, which may be in electrical, mechanical or other forms.
  • Units described as separate components may or may not be physically separated, and components shown as units may or may not be physical units, that is, may be located in one place, or may be distributed over network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution in this implementation manner.
  • each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit.
  • the above-mentioned integrated units may be implemented in the form of hardware, or may be implemented in the form of software functional units.
  • the integrated unit if implemented as a software functional unit and sold or used as a stand-alone product, may be stored in a computer-readable storage medium.
  • the technical solutions of the present application can be embodied in the form of software products in essence, or the parts that contribute to the prior art, or all or part of the technical solutions, and the computer software products are stored in a storage medium , including several instructions to make a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor (processor) to execute all or part of the steps of the methods of the various embodiments of the present application.
  • the aforementioned storage medium includes: U disk, mobile hard disk, Read-Only Memory (ROM, Read-Only Memory), Random Access Memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program codes .
  • Embodiments of the present disclosure provide a blood vessel image detection method, a detection model training method, and related devices and equipment, wherein the blood vessel image detection method includes: acquiring a three-dimensional blood vessel image of an object to be measured; In the vertical direction, at least one frame of two-dimensional image is extracted from the three-dimensional blood vessel image; the two-dimensional image is detected to obtain a regional detection result of the two-dimensional image, wherein the regional detection result includes the first sub-section corresponding to the plaque in the blood vessel. area.
  • pixels belonging to plaques in blood vessels in a three-dimensional blood vessel image can be detected, pixel-level detection can be realized, the detection accuracy of plaques in blood vessels can be improved, and the detection accuracy of plaques in blood vessels can be improved.
  • Subsequent quantification of the stenosis value of the blood vessel provides data support.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Molecular Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Medical Informatics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Quality & Reliability (AREA)
  • Apparatus For Radiation Diagnosis (AREA)

Abstract

本申请公开了一种血管图像检测方法及检测模型训练方法、相关装置、设备，其中，血管图像检测方法包括：获取待测对象的三维血管图像；沿与三维血管图像中血管的延伸方向垂直的方向，从三维血管图像中提取至少一帧二维图像；对二维图像进行检测，得到二维图像的区域检测结果，其中，区域检测结果包括与血管中的斑块对应的第一子区域。上述方案，能够提高血管中斑块检测的精度。

Description

血管图像检测方法及检测模型训练方法、相关装置、设备
相关申请的交叉引用
本申请基于申请号为202110396660.6、申请日为2021年04月13日，申请名称为“血管图像检测方法及检测模型训练方法、相关装置、设备”的中国专利申请提出，并要求该中国专利申请的优先权，该中国专利申请的全部内容在此以引入方式结合在本申请中。
技术领域
本申请涉及图像处理技术领域，特别是涉及一种血管图像检测方法及检测模型训练方法、相关装置、设备。
背景技术
随着生活水平的提高，由饮食、作息等因素而导致的血管疾病成为当今高发疾病之一。故此，有必要通过扫描等方式获取待测对象的血管图像，以便根据血管图像对待测对象的血管进行评价。
现有方式只能根据血管图像检测出血管是否存在斑块，以及对应的大致狭窄程度(如，轻度狭窄、中度狭窄、重度狭窄等)，故斑块的检测精度较低，由此而导致无法根据检测结果量化血管的狭窄程度值。有鉴于此，如何提高血管中斑块检测的精度成为亟待解决的问题。
发明内容
本申请提供一种血管图像检测方法及检测模型训练方法、相关装置、设备。
本申请第一方面提供了一种血管图像检测方法，包括：获取待测对象的三维血管图像；沿与三维血管图像中血管的延伸方向垂直的方向，从三维血管图像中提取至少一帧二维图像；对二维图像进行检测，得到二维图像的区域检测结果，其中，区域检测结果包括与血管中的斑块对应的第一子区域。
因此，通过获取待测对象的三维血管图像，并沿与三维血管图像中血管的延伸方向垂直的方向，从三维血管图像中提取出至少一帧二维图像，从而对二维图像进行检测，得到二维图像的区域检测结果，且区域检测结果包括与血管中的斑块对应的第一子区域，故能够检测出三维血管图像中属于血管中斑块的像素点，实现像素级别的检测，能够提高血管中斑块的检测精度，且能够有利于为后续量化血管的狭窄程度值提供数据支撑。
其中，区域检测结果还包括与血管对应的第一区域；在对二维图像进行检测，得到二维图像的区域检测结果之后，方法还包括：基于至少一帧二维图像中的第一区域和第一子区域，得到待测对象血管的狭窄程度值。
因此，区域检测结果还包括与血管对应的第一区域，从而基于至少一帧二维图像中的第一区域和第一子区域，得到待测对象血管的狭窄程度值，进而能够提高用户体验。
其中，基于至少一帧二维图像中的第一区域和第一子区域，得到待测对象血管的狭窄程度值，包括：利用第一子区域的面积和第一区域的第一面积，确定狭窄程度值；或者，利用第一区域在径向方向上最小的第一宽度和第一子区域在第一区域的径向方向上最小的第二宽度，确定狭窄程度值；或者，在三维血管图像包括多帧二维图像的情况下， 利用多帧二维图像中第一区域的面积与第一子区域的面积的差值，确定狭窄程度值；或者，在三维血管图像包括多帧二维图像的情况下，利用多帧二维图像中第一区域在径向方向上最小的第一宽度与第一子区域在第一区域的径向方向上最小的第二宽度的差值，确定狭窄程度值。
因此，通过利用第一子区域的面积和第一区域的第一面积，确定狭窄程度值，或者通过利用第一区域在径向方向上最小的第一宽度和第一子区域在第一区域的径向方向上最小的第二宽度，确定狭窄程度值，或者在三维血管图像包括多帧二维图像的情况下，利用多帧二维图像中第一区域的面积与第一子区域的面积的差值，确定狭窄程度值，或者在三维血管图像包括多帧二维图像的情况下，利用多帧二维图像中第一区域在径向方向上最小的第一宽度与第一子区域在第一区域的径向方向上最小的第二宽度的差值，确定狭窄程度值，能够有利于快速量化血管的狭窄程度值。
其中，三维血管图像包括多帧二维图像，对二维图像进行检测，得到二维图像的区域检测结果，包括：分别将每帧二维图像作为待测图像，并对待测图像和与待测图像间隔预设帧数内的二维图像进行检测，得到待测图像的区域检测结果。
因此，三维血管图像包括多帧二维图像，从而分别将每帧二维图像作为待测图像，并对待测图像和与待测图像间隔预设帧数内的二维图像进行检测，得到待测图像的区域检测结果，故能够利用二维图像中帧间连续信息进行检测，有利于提高图像检测的准确性。
其中，区域检测结果还包括与血管对应的第一区域；在得到待测图像的区域检测结果之后，方法还包括：按照多帧二维图像在三维血管图像中的先后顺序，将检测后的多帧二维图像进行拼接，得到与三维血管图像对应的三维检测图像；其中，三维检测图像中包含多帧二维图像的第一区域拼接得到的第二区域，以及多帧二维图像的第一子区域拼接得到的第二子区域；利用第二区域和第二子区域，确定待测对象的血管中的斑块位置。
因此，区域检测结果还包括与血管对应的第一区域，从而按照多帧二维图像在三维血管图像中的先后顺序，将检测后的多帧二维图像进行拼接，得到与三维血管图像对应的三维检测图像，且三维检测图像中包含多帧二维图像的第一区域拼接得到的第二区域，以及多帧二维图像的第一子区域拼接得到的第二子区域，进而利用第二区域和第二子区域，确定待测对象的血管中的斑块位置，故能够融合多帧二维图像的区域检测结果，对待测对象血管中的斑块进行定位，有利于提高用户体验。
其中，血管图像检测方法还包括：利用第二子区域的体积和第二区域的体积，确定待测对象血管的狭窄程度值。
因此，通过第二子区域的体积和第二区域的体积，确定待测对象血管的狭窄程度值，能够有利于快速量化血管的狭窄程度值。
其中，与待测图像间隔预设帧数内的二维图像包括以下至少一者：位于待测图像之前的二维图像，位于待测图像之后的二维图像。
因此，通过将距待测图像预设帧数范围内的二维图像设置为包括以下至少一者：位于待测图像之前的二维图像，位于待测图像之后的二维图像，从而能够利用待测图像之前的帧间连续信息，或之后的帧间连续信息，或之前和之后的帧间连续信息，进行图像检测，能够有利于提高图像检测的准确性。
其中，区域检测结果是利用检测模型对二维图像的检测得到的；和/或，对二维图像进行检测，得到二维图像的区域检测结果，包括：对二维图像进行第一检测，确定二维图像中是否存在与血管对应的第一区域；在检测到与血管对应的第一区域的情况下，对二维图像中的第一区域进行第二检测，得到第一区域内的第一子区域。
因此，区域检测结果是利用检测模型对二维图像的检测得到的，能够有利于提高图像检测的效率；而通过对二维图像进行第一检测，确定二维图像中是否存在与血管对应的第一区域，并在检测到与血管对应的第一区域的情况下，对二维图像的第一区域进行第二检测，得到第一区域内的第一子区域，能够将二维图像的检测分为血管检测、斑块检测两步，从而能够有利于提高图像检测的鲁棒性。
其中，对二维图像进行第一检测，确定二维图像中是否存在与血管对应的第一区域，包括：利用检测模型的第一检测子网络对二维图像进行第一检测，确定二维图像中是否存在与血管对应的第一区域；对二维图像中的第一区域进行第二检测，得到第一区域内的第一子区域，包括：利用检测模型的第二检测子网络对二维图像中的第一区域进行第二检测，得到第一区域内的第一子区域。
因此，通过利用检测模型的第一检测子网络对二维图像进行第一检测，确定二维图像中是否存在与血管对应的第一区域，并利用检测模型的第二检测子网络对二维图像中的第一区域进行第二检测，得到第一区域内的第一子区域，能够有利于提高图像检测的效率。
其中，第一检测子网络和第二检测子网络具有相同的网络结构；和/或，方法还包括：在未检测到与血管对应的第一区域的情况下，对二维图像的下一帧图像执行第一检测。
因此，通过将第一检测子网络和第二检测子网络设置为具有相同的网络结构，能够有利于简化检测模型的网络结构；而在未检测到与血管对应的第一区域的情况下，直接对二维图像的下一帧图像执行第一检测，能够有利于提高图像检测的效率。
本申请第二方面提供了一种检测模型的训练方法，包括：获取至少一帧样本二维图像，且至少一帧样本二维图像是从样本三维血管图像中眼与血管的延伸方向垂直的方向提取得到的，样本二维图像中标注有与血管中的斑块对应的第一实际子区域；利用检测模型对样本二维图像进行检测，得到样本二维图像中与血管中的斑块对应的第一预测子区域；利用第一实际子区域和第一预测子区域的差异，调整检测模型的网络参数。
因此，通过获取至少一帧样本二维图像，且至少一帧样本二维图像是从样本三维血管图像中眼与血管的延伸方向垂直的方向提取得到的，样本二维图像中标注有与血管中的斑块对应的第一实际子区域，并利用检测模型对样本二维图像进行检测，得到样本二维图像中与血管中的斑块对应的第一预测子区域，从而利用第一实际子区域和第一预测子区域的差异，调整检测模型的网络参数，能够使检测模型测出属于血管中斑块的像素点，从而实现像素级别的检测，能够提高血管中斑块的检测精度，且能够有利于为后续量化血管的狭窄程度值提供数据支撑。
其中，样本三维血管图像包括多帧样本二维图像；利用检测模型对样本二维图像进行检测，得到样本二维图像中与血管中的斑块对应的第一预测子区域，包括：分别将每帧样本二维图像作为样本待测图像，并利用检测模型对样本待测图像和与样本待测图像间隔预设帧数内的样本二维图像进行检测，得到样本待测图像中的第一预测子区域。
因此，至少一帧样本二维图像是从样本三维血管图像中提取得到的，且样本三维血管图像包括多帧样本二维图像，从而分别将每帧样本二维图像作为样本待测图像，并利用检测模型对样本待测图像和与样本待测图像间隔预设帧数内的样本二维图像进行检测，得到样本待测图像中的第一预测子区域，进而能够利用样本二维图像中帧间连续信息进行检测，有利于提高图像检测的准确性。
其中，用检测模型对样本二维图像进行检测，得到样本二维图像中与血管中的斑块对应的第一预测子区域，包括：利用检测模型的第一检测子网络对样本二维图像进行第一检测，确定样本二维图像中是否存在与血管对应的第一预测区域；在检测到与血管对应的第一预测区域的情况下，利用检测模型的第二检测子网络对样本二维图像中的第一 预测区域进行第二检测，得到第一预测区域内的第一预测子区域。
因此，通过利用检测模型的第一检测子网络对样本二维图像进行第一检测，确定样本二维图像中是否存在与血管对应的第一预测区域，并在检测到与血管对应的第一预测区域的情况下，利用检测模型的第二检测子网络对样本二维图像中的第一预测区域进行第二检测，得到第一预测区域内的第一预测子区域，能够将样本二维图像的检测分为血管检测、斑块检测两步，从而能够有利于提高图像检测的鲁棒性。
本申请第三方面提供了一种血管图像检测装置，包括：图像获取模块、图像提取模块和区域检测模块，图像获取模块配置为获取待测对象的三维血管图像；图像提取模块配置为沿与三维血管图像中血管的延伸方向垂直的方向，从三维血管图像中提取至少一帧二维图像；区域检测模块配置为对二维图像进行检测，得到二维图像的区域检测结果，其中，区域检测结果包括与血管中的斑块对应的第一子区域。
本申请第四方面提供了一种检测模型的训练装置，包括：样本获取模块、区域检测模块和参数调整模块，样本获取模块配置为获取至少一帧样本二维图像，且至少一帧样本二维图像是从样本三维血管图像中眼与血管的延伸方向垂直的方向提取得到的，且样本二维图像中标注有与血管中的斑块对应的第一实际子区域；区域检测模块配置为利用检测模型对样本二维图像进行检测，得到样本二维图像中与血管中的斑块对应的第一预测子区域；参数调整模块配置为利用第一实际子区域和第一预测子区域的差异，调整检测模型的网络参数。
本申请第五方面提供了一种电子设备，包括相互耦接的存储器和处理器，处理器配置为执行存储器中存储的程序指令，以实现上述第一方面中的血管图像检测方法，或实现上述第二方面中的检测模型的训练方法。
本申请第六方面提供了一种计算机可读存储介质，其上存储有程序指令，程序指令被处理器执行时实现上述第一方面中的血管图像检测方法，或实现上述第二方面中的检测模型的训练方法。
上述方案，通过获取待测对象的三维血管图像，并沿与三维血管图像中血管的延伸方向垂直的方向，从三维血管图像中提取出至少一帧二维图像，从而对二维图像进行检测，得到二维图像的区域检测结果，且区域检测结果包括与血管中的斑块对应的第一子区域，故能够检测出三维血管图像中属于血管中斑块的像素点，实现像素级别的检测，能够提高血管中斑块的检测精度，且能够有利于为后续量化血管的狭窄程度值提供数据支撑。
附图说明
图1是本申请血管图像检测方法一实施例的流程示意图；
图2是对二维图像进行检测一实施例的示意图；
图3是三维检测图像一实施例的示意图；
图4是本申请血管图像检测方法另一实施例的流程示意图；
图5是本申请检测模型的训练方法一实施例的流程示意图；
图6是本申请血管图像检测装置一实施例的结构示意图；
图7是本申请检测模型的训练装置一实施例的结构示意图；
图8是本申请电子设备一实施例的结构示意图；
图9是本申请计算机可读存储介质一实施例的结构示意图。
具体实施方式
下面结合说明书附图，对本申请实施例的方案进行详细说明。
以下描述中，为了说明而不是为了限定，提出了诸如特定系统结构、接口、技术之类的具体细节，以便透彻理解本申请。
本文中术语“系统”和“网络”在本文中常被可互换使用。本文中术语“和/或”，仅仅是一种描述关联对象的关联关系，表示可以存在三种关系，例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B这三种情况。另外，本文中字符“/”，一般表示前后关联对象是一种“或”的关系。此外，本文中的“多”表示两个或者多于两个。
本公开实施例提供了一种血管图像检测方法，其执行主体可以是血管图像检测装置，例如，血管图像检测方法可以由终端设备或服务器或其它电子设备执行，其中，终端设备可以为用户设备(User Equipment，UE)、移动设备、用户终端、终端、蜂窝电话、无绳电话、个人数字助理(Personal Digital Assistant，PDA)、手持设备、计算设备、车载设备、可穿戴设备等。在一些可能的实现方式中，该血管图像检测方法可以通过处理器调用存储器中存储的计算机可读指令的方式来实现。
请参阅图1，图1是本申请血管图像检测方法一实施例的流程示意图。具体而言，可以包括如下步骤：
步骤S11：获取待测对象的三维血管图像。
本公开实施例中，可以通过扫描待测对象预设部位的血管，从而得到待测对象的三维血管图像。预设部位可以根据实际应用需要进行设置，具体可以包括但不限于：心脏冠状动脉、颈内动脉、颈外动脉、椎动脉等等，在此不做限定。由于血管通常行径扭曲，不处于同一平面上，故扫描得到的医学影像(如，计算机断层扫描等)可以利用曲面重建(Curved Planar Reformation，CPR)进行拉直，得到拉直的CPR图像，并将其作为待测对象的三维血管图像。具体地，可以对扫描得到的三维血管图像进行初始检测，以得到三维血管图像中的血管，例如，可以采用神经网络(如，U-Net等)、传统图像分割方式(如，边缘提取，以及膨胀腐蚀等形态学处理方式)，具体在此不再赘述。从而可以确定血管的中心线，并沿血管的中心线，提取血管的横截面图像，并将属于同一血管的横截面图像依序堆叠，即可得到本公开实施例中待测对象的三维血管图像。
步骤S12：沿与三维血管图像中血管的延伸方向垂直的方向，从三维血管图像中提取至少一帧二维图像。
本公开实施例中，血管的延伸方向是指血管中心线的切线方向。具体地，可以在血管的各个位置处沿其中心线法平面，提取二维图像。具体地，对于尺寸为H(高度)*W(宽度)*D(深度)的三维血管图像，可以在血管的各个位置处沿其中心线法平面，提取一待处理图像，并将待处理图像经过插值、裁剪等过程，得到预设尺寸(如，64*64)的二维图像。
步骤S13：对二维图像进行检测，得到二维图像的区域检测结果，其中，区域检测结果包括与血管中的斑块对应的第一子区域。
在一个实施场景中，可以对每帧二维图像进行单独检测，得到二维图像的区域检测结果。例如，可以对二维图像进行边缘检测，得到包含边缘信息(如，血管边缘信息、斑块边缘信息等)的掩膜(mask)图像，并对掩膜图像进行膨胀腐蚀等形态学处理，滤除掩膜图像中的离群点等异常像素点，从而可以得到二维图像的区域检测结果。
在另一个实施场景中，三维血管图像可以提取得到多帧二维图像，例如，50帧二维图像、100帧二维图像、150帧二维图像等等，在此不做限定。则还可以分别将每帧二维图像作为待测图像，并对待测图像和与待测图像间隔预设帧数内的二维图像进行检测，得到待测图像的区域检测结果。上述方式，可以利用二维图像中的帧间连续信息进行检 测，有利于提高图像检测的准确性。
在一个具体的实施场景中，预设帧数可以根据实际应用需要进行设置，例如：1、2、3、4等等，在此不做限定。此外，当对每帧二维图像进行单独检测时，可以认为预设帧数为0，即与待测图像间隔预设帧数内的二维图像即为待测图像本身。
在另一个具体的实施场景中，与待测图像间隔预设帧数内的二维图像可以包括以下至少一者：位于待测图像之前的二维图像，位于待测图像之后的二维图像。即与待测图像间隔预设帧数内的二维图像可以是位于待测图像之前且与待测图像间隔预设帧数内的二维图像，例如，位于待测图像之前3帧内的二维图像；或者，与待测图像间隔预设帧数内的二维图像也可以是位于待测图像之后且与待测图像间隔预设帧数内的二维图像，例如，位于待测图像之后3帧内的二维图像；或者，与待测图像间隔预设帧数内的二维图像也可以包括：位于待测图像之前且与待测图像间隔预设帧数内的二维图像和位于待测图像之后且与待测图像间隔预设帧数内的二维图像，例如，位于待测图像之前3帧内的二维图像和位于待测图像之后3帧内的二维图像。
在又一个实施场景中，请结合参阅图2，图2是对二维图像进行检测一实施例的示意图，为了提高检测效率，还可以预先训练一检测模型，从而可以利用检测模型对二维图像进行检测，得到二维图像的区域检测结果。检测模型具体可以采用U-Net、全卷积网络(Fully Convolutional Network，FCN)等等，在此不做限定。检测模型的训练步骤可以参阅后续公开实施例中的步骤，在此暂不赘述。此外，检测模型中用于提取图像特征的卷积核具体可以设置为二维卷积核，则可以对每帧二维图像进行单独检测，得到二维图像的区域检测结果，也可以分别将每帧二维图像作为待测图像，并对待测图像和与待测图像间隔预设帧数内的二维图像进行检测，得到待测图像的区域检测结果。或者，检测模型中用于提取图像特征的卷积核具体可以设置为三维卷积核，则可以分别将每帧二维图像作为待测图像，并对待测图像和与待测图像间隔预设帧数内的二维图像进行检测，得到待测图像和与待测图像间隔预设帧数内的二维图像的区域检测结果。具体可以根据实际应用需要进行设置，在此不做限定。
在一个具体的实施场景中，可以利用检测模型分别对各二维图像进行检测，得到二维图像的区域检测结果。
在另一个具体的实施场景中，还可以分别将每帧二维图像作为待测图像，并利用检测模型对待测图像和与待测图像间隔预设帧数内的二维图像进行检测，得到待测图像的区域检测结果，故能够利用二维图像中的帧间连续信息进行检测，有利于提高图像检测的准确性。与待测图像间隔预设帧数内的二维图像具体的设置方式，可以参阅前述描述，在此不再赘述。
在又一个实施场景中，区域检测结果还可以包括与血管对应的第一区域，如图2所示，斜线填充部分为与血管对应的第一区域，第一区域中白色填充部分为与斑块对应的第一子区域，则可以基于至少一帧二维图像中第一区域和第一子区域，得到待测对象血管的狭窄程度值。上述方式，可以根据检测得到的第一子区域和第一区域，对血管的狭窄程度进行快速统计、量化，有利于提高用户体验。
在一个具体的实施场景中，可以利用第一子区域的面积和第一区域的面积，确定狭窄程度值。例如，可以将检测得到的第一子区域的面积与第一区域的面积之间的比值，作为狭窄程度值，在此情况下，狭窄程度值越大，表明斑块对血管的堵塞程度越高，血管狭窄程度越高，反之，狭窄程度值越小，表明斑块对血管的堵塞程度越低，血管狭窄程度越低。
在另一个具体的实施场景中，还可以利用第一区域在径向方向上最小的第一宽度D和第一子区域在第一区域的径向方向上最小的第二宽度d，确定狭窄程度值。需要说明 的是，血管中心线在二维图像中经过一像素点(如图2中以黑色填充圆形所示)，则在二维图像中经过该像素点的方向即为径向方向。请结合参阅图2，斑块的第一子区域在第一区域的某一径向方向(图2中双向箭头所示)上的第二宽度为d1，第一区域在该径向方向上的第一宽度为D1，此外，斑块的第一子区域在第一区域的另一径向方向(图2中双向箭头所示)上的第二宽度为d2，第一区域在该径向方向上的第一宽度为D2，以此类推，在不同径向方向可以分别得到斑块的第一区域的第二宽度以及第一区域的第一宽度，通过比较可以得到最小的第二宽度d和最小的第一宽度D，则可以利用第二宽度d与第一宽度D的比值作为该径向方向上的狭窄程度值，在此情况下，狭窄程度值越大，表明斑块对血管的堵塞程度越高，血管狭窄程度越高，反之，狭窄程度值越小，表明斑块对血管的堵塞程度越低，血管狭窄程度越低。
在又一个具体的实施场景中，还可以利用其它方式计算狭窄程度值，例如，针对椎-基底动脉等情况，可以采用NASCET(North American Symptomatic Carotid Endarterectomy，北美症状性颈动脉内膜切除试验)法、ECST(MRC European Carotid Surgery Trial，欧洲颈动脉外科试验)法等计算，其它场景可以根据实际应用情况进行设置，在此不再一一举例。
在又一个具体的实施场景中，在三维血管图像包括多帧二维图像的情况下，还可以利用多帧二维图像中第一区域的面积与第一子区域的面积的差值，确定狭窄程度值。具体地，可以获取每帧二维图像中第一区域的面积与S与第一子区域的面积s的差值S-s，作为二维图像中血管的有效面积，为了便于描述，可以记为S′。当血管中不存在斑块，或仅存在微小斑块时，多帧二维图像中血管的有效面积S′应随血管的延伸而稳定变化，反之，当血管中存在较大尺寸斑块时，多帧二维图像中血管的有效面积S′应随血管的延伸而发生突变。在此情况下，可以利用预设方式对多帧二维图像中血管的有效面积S′进行拟合，得到有效面积S′的理论值，并将实际测量到的血管的有效面积S′的值，作为血管的有效面积S′的实际值，从而可以将实际值偏离理论值的百分比，作为血管的狭窄程度值，则计算得到的狭窄程度值越大，表明血管越狭窄；或者，也可以直接将有效面积S′的实际值与有效面积S′的理论值的比值，作为血管的狭窄程度值，则计算得到的狭窄程度值越小，表明血管越狭窄。
在又一个具体的实施场景中，在三维血管图像包括多帧二维图像的情况下，还可以利用多帧二维图像中第一区域在径向方向上最小的第一宽度与第一子区域在第一区域的径向方向上最小的第二宽度的差值，确定狭窄程度值。具体地，可以获取每帧二维图像中第一子区域在第一区域的径向方向上最小的第二宽度d和第一区域最小的第一宽度D，并将血管的第一宽度D与斑块的第二宽度d的差值D-d，作为二维图像中血管的有效宽度，为了便于描述，可以记为D′。当血管中不存在斑块，或仅存在微小斑块时，多帧二维图像中血管的有效宽度D′应随血管的延伸呈稳定变化(如，线性增加，或线性减小)，例如，第一帧二维图像中血管的有效宽度D′为2mm，第二帧二维图像中血管的有效宽度D′为1.9mm，第三帧二维图像中血管的有效宽度D′为1.8mm，通过统计，多帧二维图像中血管的有效宽度D′以趋近于D′＝2-0.1*(x-1)的线性关系变化，其中，x表示第x帧二维图像；而当血管中存在较大尺寸斑块时，多帧二维图像中血管的有效宽度D′应随血管的延伸呈非线性变化，例如，第一帧二维图像中血管的有效宽度D′为2mm，第二帧二维图像中血管的有效宽度D′为1.9mm，第三帧二维图像中血管的有效宽度D′为1.6mm，第四帧二维图像中血管的有效宽度D′为1.3mm，第五帧二维图像中血管的有效宽度D′为0.9mm，由此可见，有效宽度D′随血管的延伸呈现出突变。在此情况下，可以利用预设方式对多帧二维图像中血管的有效宽度D′进行线性拟合，得到 有效宽度D′随血管的延伸呈线性变化的值，作为血管的有效宽度D′的理论值，并将实际测量到的血管的有效宽度D′的值，作为血管的有效宽度D′的实际值，从而可以将实际值偏离理论值的百分比，作为血管的狭窄程度值，则计算得到的狭窄程度值越大，表明血管越狭窄；或者，也可以直接将有效宽度D′的实际值与有效宽度D′的理论值的比值，作为狭窄程度值，则计算得到的狭窄程度值越小，表明血管越狭窄。
在又一个具体的实施场景中，在得到待测对象血管的狭窄程度值之后，还可以利用狭窄程度值生成检测报告。例如，检测报告可以以预设格式设置，则可以将检测得到的狭窄程度值添加至检测报告中与狭窄程度值对应的字段中。此外，还可以将标注有区域检测结果的图像作为附件附于检测报告。
在又一个实施场景中，区域检测结果还可以进一步包括与血管对应的第一区域，则还可以按照多帧二维图像在三维血管图像中的先后顺序，将检测后的多帧二维图像进行拼接，得到与三维血管图像对应的三维检测图像，且三维检测图像中包含多帧二维图像的第一区域拼接得到的第二区域以及多帧二维图像的第一子区域拼接得到的第二子区域，从而可以利用第二区域和第二子区域，确定待测对象的血管中的斑块位置。上述方式，可以融合多帧二维图像的区域检测结果，对待测对象血管中的斑块进行定位，有利于提高用户体验。
在一个实施场景中，在得到三维检测图像之后，可以进一步利用第二子区域的体积和第二区域的体积，确定狭窄程度值。具体地，第二区域的体积可以是第二区域所包含的像素点个数，类似地，第二子区域的体积可以是第二子区域所包含的像素点个数。
请结合参阅图3，图3是三维检测图像一实施例的示意图，如图3所示，长方体区域表示多帧二维图像拼接形成的三维检测图像，长方体区域内圆柱域表示第一区域拼接形成的与血管对应的第二区域，小点填充区域表示第一子区域拼接形成的与斑块对应的第二子区域，通过融合多帧二维图像的区域检测结果，能够实现对斑块的快速定位，提高用户体验。
在又一个实施场景中，区域检测结果还可以包括检测得到的斑块的具体类型，如钙化斑块、混合斑块、非钙化斑块等等，在此不做限定。
上述方案，通过获取待测对象的三维血管图像，并沿垂直于三维血管图像中血管的延伸方向的方向，从三维血管图像中提取出至少一帧二维图像，从而对二维图像进行检测，得到二维图像的区域检测结果，且区域检测结果包括与血管中的斑块对应的第一子区域，故能够检测出三维血管图像中属于血管中斑块的像素点，实现像素级别的检测，能够提高血管中斑块的检测精度，且能够有利于为后续量化血管的狭窄程度值提供数据支撑。
请参阅图4，图4本申请血管图像检测方法另一实施例的流程示意图。具体可以包括如下步骤：
步骤S41：获取待测对象的三维血管图像。
具体可以参阅前述公开实施例中的相关步骤，在此不再赘述。
步骤S42：沿垂直于三维血管图像中血管的延伸方向的方向，从三维血管图像中提取至少一帧二维图像。
具体可以参阅前述公开实施例中的相关步骤，在此不再赘述。
步骤S43：对二维图像进行第一检测，确定二维图像中是否存在与血管对应的第一区域。
在一个实施场景中，可以对二维图像进行边缘检测，得到包含边缘信息的掩膜(mask)图像，并对掩膜图像进行膨胀腐蚀等形态学处理，滤除掩膜图像中的离群点等异常像素点，从而可以得到二维图像中边缘线，由边缘线封闭而成的区域，可以作为与边缘线对 应的封闭区域，进而可以将封闭区域最大的边缘线作为血管的边缘线，并将该封闭区域作为与血管对应的第一区域。
在另一个实施场景中，还可以利用检测模型的第一检测子网络对二维图像进行第一检测，确定二维图像中是否存在与血管对应的第一区域，从而可以通过检测模型进行检测，提高检测效率。检测模型的具体结构可以参阅前述公开实施例中的相关描述，在此不再赘述。
在又一个实施场景中，还可以分别将每帧二维图像作为待测图像，并利用待测图像和与待测图像间隔预设帧数内的二维图像进行检测，确定二维图像中是否存在与血管对应的第一区域。预设帧数的具体设置方式可以参阅前述公开实施例中的相关描述，在此不再赘述。具体地，可以利用检测模型的第一检测子网络对待测图像和与待测图像间隔预设帧数内的二维图像进行第一检测，确定二维图像中是否存在与血管对应的第一区域。检测模型中用于提取图像特征的卷积核的具体设置方式，可以参阅前述公开实施例中的相关描述，在此不再赘述。
步骤S44：在检测到与血管对应的第一区域的情况下，对二维图像中的第一区域进行第二检测，得到第一区域内的第一子区域。
本公开实施例中，在检测到与血管对应的第一区域的情况下，对二维图像的第一区域进行第二检测，得到第一区域内的第一子区域，即与斑块对应的第一子区域。
在一个实施场景中，可以利用边缘检测的方式对第一区域进行第二检测，从而得到第一区域内与斑块对应的第一子区域，具体可以参阅前述利用边缘检测得到与血管对应的第一区域的相关描述，在此不再赘述。
在另一个实施场景中，为了提高检测效率，还可以利用检测模型的第二检测子网络对二维图像中的第一区域进行第二检测，得到第一区域内的第一子区域。
在一个具体的实施场景中，上述第一检测子网络和第二检测子网络可以具有相同的网络结构，从而可以简化检测模型的网络结构。例如，第一检测子网络和第二检测子网络可以均为U-Net，即检测模型是由两个串联的U-Net所组成的，其他场景可以根据实际应用需要进行设置，在此不做限定。
在另一个实施场景中，在未检测到与血管对应的第一区域的情况下，可以对二维图像的下一帧图像执行第一检测，具体可以参阅前述相关描述，在此不再赘述。上述方式，可以在某一帧二维图像中未检测到与血管对应的第一区域的情况下，直接对下一帧二维图像执行第一检测，以检测下一帧二维图像中是否存在与血管对应的第一区域，从而可以提高检测效率。
区别于前述实施例，通过对二维图像进行第一检测，确定二维图像中是否存在与血管对应的第一区域，并在检测到与血管对应的第一区域的情况下，对二维图像的第一区域进行第二检测，得到第一区域内的第一子区域，能够将二维图像的检测分为血管检测、斑块检测两步，从而能够有利于提高图像检测的鲁棒性。
请参阅图5，图5是本申请图像检测模型的训练方法一实施例的流程示意图。具体可以包括如下步骤；
步骤S51：获取至少一帧样本二维图像，至少一帧样本二维图像是从样本三维血管图像中沿与血管的延伸方向垂直的方向提取得到的，且样本二维图像中标注有与血管中的斑块对应的第一实际子区域。
在一个实施场景中，样本二维图像具体可以是从样本三维血管图像中沿与血管的延伸方向垂直的方向提取得到的。样本三维血管图像具体获取方式可以参阅前述公开实施例中的相关描述，在此不再赘述。此外，具体的提取方式可以参阅前述公开实施例中关于从三维血管图像中提取至少一帧二维图像的方式，在此不再赘述。
在一个实施场景中，样本二维图像中还可以标注有斑块的具体类型，如钙化斑块、混合斑块、非钙化斑块等等，在此不做限定。
在一个实施场景中，样本二维图像中还可以标注有与血管对应的第一实际区域。
步骤S52：利用检测模型对样本二维图像进行检测，得到样本二维图像中与血管中的斑块对应的第一预测子区域。
本公开实施例中，检测模型具体可以根据实际应用需要进行设置，具体可以参阅前述公开实施例中的相关描述，在此不再赘述。
在一个实施场景中，检测模型还可以对样本二维图像进行检测，得到样本二维图像中与血管对应的第一预测区域。
在一个具体的实施场景中，检测模型具体可以包括第一检测子网络和第二检测子网络，从而可以利用检测模型的第一检测子网络对样本二维图像进行第一检测，确定样本二维图像中是否存在与血管对应的第一预测区域，并在检测到与血管对应的第一预测区域的情况下，利用检测模型的第二检测子网络对样本二维图像中的第一预测区域进行第二检测，得到第一预测区域内的第一预测子区域，从而能够有利于提高检测效率。具体地，第一检测子网络和第二检测子网络可以具有相同的网络结构，例如，第一检测子网络和第二检测子网络均是U-Net，即检测模型是由串联的U-Net组成的，其他场景也可以根据实际应用需要进行设置，在此不做限定。此外，在未检测到与血管对应的第一预测区域的情况下，还可以利用检测模型的第二检测子网络对下一帧样本二维图像执行上述第一检测，具体可以参阅前述公开实施例中的相关描述，在此不再赘述。
在一个实施场景中，上述样本三维血管图像包括多帧样本二维图像，则可以分别将每帧样本二维图像作为样本待测图像，并利用检测模型对样本待测图像和与样本待测图像间隔预设帧数内的样本二维图像进行检测，得到样本待测图像中的第一预测子区域，从而可以融合帧间连续信息进行检测，有利于提高检测准确性。与样本待测图像间隔预设帧数内的样本二维图像的具体设置方式，可以参阅前述公开实施例中关于与待测图像间隔预设帧数内的二维图像的设置方式，在此不再赘述。
在另一个实施场景中，可以利用检测模型的第一检测子网络对样本待测图像和与样本待测图像间隔预设帧数内的样本二维图像进行检测，先确定样本待测图像中是否存在与血管对应的第一预测区域，并在检测到与血管对应的第一预测区域的情况下，利用检测模型的第二检测子网络对样本二维图像中的第一预测区域进行第二检测，得到第一预测区域内的第一预测子区域。上述方式，可以在检测时融合帧间连续信息，并将图像检测分为血管检测、斑块检测两部分，从而能够在提高图像检测准确性的同时，提高图像检测的效率。
步骤S53：利用第一实际子区域和第一预测子区域的差异，调整检测模型的网络参数。
在一个实施场景中，可以利用预设损失函数(如交叉熵损失函数、dice loss函数)处理第一实际子区域和第一预测子区域，得到检测模型的第一损失值，从而利用第一损失值调整检测模型的网络参数。此外，在样本二维图像中还标注有与血管对应的第一实际区域，检测模型还检测得到与血管对应的第一预测区域的情况下，还可以利用预设损失函数(如交叉熵损失函数、dice loss函数)处理第一实际区域和第一预测区域，得到检测模型的第二损失值，从而利用第一损失值和第二损失值调整检测模型的网络参数。此外，在样本二维图像中还标注有斑块的实际类型，检测模型还检测得到斑块的预测类型，还可以利用预设损失函数(如交叉熵损失函数)处理实际类型和预测类型，得到检测模型的第三损失值，从而可以利用上述第一损失值、第二损失值和第三损失值调整检测模型的网络参数。
在一个具体的实施场景中，预设损失函数可以包括但不限于：交叉熵损失函数(cross entropy loss)、dice loss等等，在此不做限定。
在另一个具体的实施场景中，具体可以采用随机梯度下降(Stochastic Gradient Descent，SGD)、批量梯度下降(Batch Gradient Descent，BGD)、小批量梯度下降(Mini-Batch Gradient Descent，MBGD)等方式，利用上述损失值对检测模型的网络参数进行调整，其中，批量梯度下降是指在每一次迭代时，使用所有样本来进行参数更新；随机梯度下降是指在每一次迭代时，使用一个样本来进行参数更新；小批量梯度下降是指在每一次迭代时，使用一批样本来进行参数更新，在此不再赘述。
在又一个具体的实施场景中，还可以设置一训练结束条件，当满足训练结束条件时，可以结束对检测模型的训练。具体地，训练结束条件可以包括：损失值小于一预设损失阈值，且损失值不再减小；当前训练次数达到预设次数阈值(例如，500次、1000次等)，在此不做限定。
上述方案，通过获取至少一帧样本二维图像，且样本二维图像中标注有与血管中的斑块对应的第一实际子区域，并利用检测模型对样本二维图像进行检测，得到样本二维图像中与血管中的斑块对应的第一预测子区域，从而利用第一实际子区域和第一预测子区域的差异，调整检测模型的网络参数，能够使检测模型测出属于血管中斑块的像素点，从而实现像素级别的检测，能够提高血管中斑块的检测精度，且能够有利于为后续量化血管的狭窄程度值提供数据支撑。
请参阅图6，图6是本申请血管图像检测装置60一实施例的结构示意图。血管图像检测装置60包括：图像获取模块61、图像提取模块62和区域检测模块63，图像获取模块61配置为获取待测对象的三维血管图像；图像提取模块62配置为沿与三维血管图像中血管的延伸方向垂直的方向，从三维血管图像中提取至少一帧二维图像；区域检测模块63配置为对二维图像进行检测，得到二维图像的区域检测结果，其中，区域检测结果包括与血管中的斑块对应的第一子区域。
上述方案，通过获取待测对象的三维血管图像，并沿与三维血管图像中血管的延伸方向垂直的方向，从三维血管图像中提取出至少一帧二维图像，从而对二维图像进行检测，得到二维图像的区域检测结果，且区域检测结果包括与血管中的斑块对应的第一子区域，故能够检测出三维血管图像中属于血管中斑块的像素点，实现像素级别的检测，能够提高血管中斑块的检测精度，且能够为后续量化血管的狭窄程度值提供数据支撑。
在一些公开实施例中，区域检测结果还包括与血管对应的第一区域，血管图像检测装置60还包括狭窄计算模块，配置为基于至少一帧二维图像中的第一区域和第一子区域，得到待测对象血管的狭窄程度值。
区别于前述实施例，区域检测结果还包括与血管对应的第一区域，从而基于至少一帧二维图像中的第一区域和第一子区域，得到待测对象血管的狭窄程度值，进而能够提高用户体验。
在一些公开实施例中，狭窄计算模块配置为利用第一子区域的面积和第一区域的面积，确定狭窄程度值；或者，利用第一区域在径向方向上最小的第一宽度和第一子区域在第一区域的径向方向上的最小第二宽度，确定狭窄程度值；或者，在三维血管图像包括多帧二维图像的情况下，利用多帧二维图像中第一区域的面积与第一子区域的面积的差值，确定狭窄程度值；或者，在三维血管图像包括多帧二维图像的情况下，利用多帧二维图像中第一区域在径向方向上最小的第一宽度与第一子区域在第一区域的径向方向上最小的第二宽度的差值，确定狭窄程度值。
区别于前述实施例，通过利用第一子区域的面积和第一区域的第一宽度，确定狭窄程度值，或者通过利用第一子区域在第一区域的径向方向上最大的第二宽度和第一区域 的第一宽度，确定狭窄程度值，或者在三维血管图像包括多帧二维图像的情况下，利用多帧二维图像中第一区域的面积与第一子区域的面积的差值，确定狭窄程度值，或者在三维血管图像包括多帧二维图像的情况下，利用多帧二维图像中第一区域在径向方向上最小的第一宽度与第一子区域在第一区域的径向方向上最小的第二宽度的差值，确定狭窄程度值，能够有利于快速量化血管的狭窄程度值。
在一些公开实施例中，三维血管图像包括多帧二维图像，区域检测模块63具体配置为分别将每帧二维图像作为待测图像，并对待测图像和与待测图像间隔预设帧数内的二维图像进行检测，得到待测图像的区域检测结果。
区别于前述实施例，三维血管图像包括多帧二维图像，从而分别将每帧二维图像作为待测图像，并对待测图像和与待测图像间隔预设帧数内的二维图像进行检测，得到待测图像的区域检测结果，故能够利用二维图像中帧间连续信息进行检测，有利于提高图像检测的准确性。
在一些公开实施例中，区域检测结果还包括与血管对应的第一区域，血管图像检测装置60还包括图像拼接模块，配置为按照多帧二维图像在三维血管图像中的先后顺序，将检测后的多帧二维图像进行拼接，得到与三维血管图像对应的三维检测图像；其中，三维检测图像中包含多帧二维图像的第一区域拼接得到的第二区域，以及多帧二维图像的第一子区域拼接得到的第二子区域，血管图像检测装置60还包括斑块定位模块，配置为利用第二区域和第二子区域，确定待测对象的血管中的斑块位置。
区别于前述实施例，区域检测结果还包括与血管对应的第一区域，从而按照多帧二维图像在三维血管图像中的先后顺序，将检测后的多帧二维图像进行拼接，得到与三维血管图像对应的三维检测图像，且三维检测图像中包含多帧二维图像的第一区域拼接得到的第二区域，以及多帧二维图像的第一子区域拼接得到的第二子区域，进而利用第二区域和第二子区域，确定待测对象的血管中的斑块位置，故能够融合多帧二维图像的区域检测结果，对待测对象血管中的斑块进行定位，有利于提高用户体验。
在一些公开实施例中，狭窄计算模块配置为利用第二子区域的体积和第二区域的体积，确定待测对象血管的狭窄程度值。
区别于前述实施例，通过第二子区域的体积和第二区域的体积，确定待测对象血管的狭窄程度值，能够有利于快速量化血管的狭窄程度值。
在一些公开实施例中，距待测图像预设帧数内的二维图像包括以下至少一者：位于待测图像之前的二维图像，位于待测图像之后的二维图像。
区别于前述实施例，通过将距待测图像预设帧数范围内的二维图像设置为包括以下至少一者：位于待测图像之前的二维图像，位于待测图像之后的二维图像，从而能够利用待测图像之前的帧间连续信息，或之后的帧间连续信息，或之前和之后的帧间连续信息，进行图像检测，能够有利于提高图像检测的准确性。
在一些公开实施例中，区域检测结果是利用检测模型对二维图像的检测得到的；和/或，区域检测模块63包括第一检测子模块，配置为对二维图像进行第一检测，确定二维图像中是否存在与血管对应的第一区域，区域检测模块63包括第二检测子模块，配置为在检测到与血管对应的第一区域的情况下，对二维图像中的第一区域进行第二检测，得到第一区域内的第一子区域。
区别于前述实施例，区域检测结果是利用检测模型对二维图像的检测得到的，能够有利于提高图像检测的效率；而通过对二维图像进行第一检测，确定二维图像中是否存在与血管对应的第一区域，并在检测到与血管对应的第一区域的情况下，对二维图像的第一区域进行第二检测，得到第一区域内的第一子区域，能够将二维图像的检测分为血管检测、斑块检测两步，从而能够有利于提高图像检测的鲁棒性。
在一些公开实施例中，第一检测子模块具体配置为利用检测模型的第一检测子网络对二维图像进行第一检测，确定二维图像中是否存在与血管对应的第一区域，第二检测子模块具体配置为利用检测模型的第二检测子网络对二维图像中的第一区域进行第二检测，得到第一区域内的第一子区域。
区别于前述实施例，通过利用检测模型的第一检测子网络对二维图像进行第一检测，确定二维图像中是否存在与血管对应的第一区域，并利用检测模型的第二检测子网络对二维图像中的第一区域进行第二检测，得到第一区域内的第一子区域，能够有利于提高图像检测的效率。
在一些公开实施例中，第一检测子网络和第二检测子网络具有相同的网络结构，和/或，第一检测子模块还配置为在未检测到与血管对应的第一区域的情况下，对二维图像的下一帧图像执行第一检测。
区别于前述实施例，通过将第一检测子网络和第二检测子网络设置为具有相同的网络结构，能够有利于简化检测模型的网络结构；而在未检测到与血管对应的第一区域的情况下，直接对二维图像的下一帧图像执行第一检测，能够有利于提高图像检测的效率。
请参阅图7，图7是本申请检测模型的训练装置70一实施例的结构示意图。检测模型的训练装置70包括：样本获取模块71、区域检测模块72和参数调整模块73，样本获取模块71配置为获取至少一帧样本二维图像，至少一帧样本二维图像是从样本三维血管图像中沿与血管的延伸方向垂直的方向提取得到的，且样本二维图像中标注有与血管中的斑块对应的第一实际子区域；区域检测模块72配置为利用检测模型对样本二维图像进行检测，得到样本二维图像中与血管中的斑块对应的第一预测子区域；参数调整模块73配置为利用第一实际子区域和第一预测子区域的差异，调整检测模型的网络参数。
上述方案，通过获取至少一帧样本二维图像，至少一帧样本二维图像是从样本三维血管图像中沿与血管的延伸方向垂直的方向提取得到的，且样本二维图像中标注有与血管中的斑块对应的第一实际子区域，并利用检测模型对样本二维图像进行检测，得到样本二维图像中与血管中的斑块对应的第一预测子区域，从而利用第一实际子区域和第一预测子区域的差异，调整检测模型的网络参数，能够使检测模型测出属于血管中斑块的像素点，从而实现像素级别的检测，提高血管中斑块的检测精度，且能够有利于为后续量化血管的狭窄程度值提供数据支撑。
在一些公开实施例中，样本三维血管图像包括多帧样本二维图像，区域检测模块72具体配置为分别将每帧样本二维图像作为样本待测图像，并利用检测模型对样本待测图像和与样本待测图像间隔预设帧数内的样本二维图像进行检测，得到样本待测图像中的第一预测子区域。
区别于前述实施例，样本三维血管图像包括多帧样本二维图像，从而分别将每帧样本二维图像作为样本待测图像，并利用检测模型对样本待测图像和与样本待测图像间隔预设帧数内的样本二维图像进行检测，得到样本待测图像中的第一预测子区域，进而能够利用样本二维图像中帧间连续信息进行检测，有利于提高图像检测的准确性。
在一些公开实施例中，区域检测模块72包括第一检测子模块，配置为利用检测模型的第一检测子网络对样本二维图像进行第一检测，确定样本二维图像中是否存在与血管对应的第一预测区域，区域检测模块72包括第二检测子模块，配置为在检测到与血管对应的第一预测区域的情况下，利用检测模型的第二检测子网络对样本二维图像中的第一预测区域进行第二检测，得到第一预测区域内的第一预测子区域。
区别于前述实施例，通过利用检测模型的第一检测子网络对样本二维图像进行第一检测，确定样本二维图像中是否存在与血管对应的第一预测区域，并在检测到与血管对 应的第一预测区域的情况下，利用检测模型的第二检测子网络对样本二维图像中的第一预测区域进行第二检测，得到第一预测区域内的第一预测子区域，能够将样本二维图像的检测分为血管检测、斑块检测两步，从而能够有利于提高图像检测的鲁棒性。
请参阅图8，图8是本申请电子设备80一实施例的结构示意图。电子设备80包括相互耦接的存储器81和处理器82，处理器82配置为执行存储器81中存储的程序指令，以实现上述任一血管图像检测方法实施例的步骤，或实现上述任一检测模型的训练方法实施例中的步骤。在一个具体的实施场景中，电子设备80可以包括但不限于：微型计算机、服务器，此外，电子设备80还可以包括笔记本电脑、平板电脑等移动设备，在此不做限定。
具体而言，处理器82配置为控制其自身以及存储器81以实现上述任一血管图像检测方法实施例的步骤，或实现上述任一检测模型的训练方法实施例中的步骤。处理器82还可以称为CPU(Central Processing Unit，中央处理单元)。处理器82可能是一种集成电路芯片，具有信号的处理能力。处理器82还可以是通用处理器、数字信号处理器(Digital Signal Processor,DSP)、专用集成电路(Application Specific Integrated Circuit,ASIC)、现场可编程门阵列(Field-Programmable Gate Array,FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。另外，处理器82可以由集成电路芯片共同实现。
上述方案，能够提高血管中斑块的检测精度。
请参阅图9，图9为本申请计算机可读存储介质90一实施例的结构示意图。计算机可读存储介质90存储有能够被处理器运行的程序指令901，程序指令901用于实现上述任一血管图像检测方法实施例的步骤，或实现上述任一检测模型的训练方法实施例中的步骤。
上述方案，能够提高血管中斑块的检测精度。
在一些实施例中，本公开实施例提供的装置具有的功能或包含的模块可以用于执行上文方法实施例描述的方法，其具体实现可以参照上文方法实施例的描述，为了简洁，这里不再赘述。
上文对各个实施例的描述倾向于强调各个实施例之间的不同之处，其相同或相似之处可以互相参考，为了简洁，本文不再赘述。
在本申请所提供的几个实施例中，应该理解到，所揭露的方法和装置，可以通过其它的方式实现。例如，以上所描述的装置实施方式仅仅是示意性的，例如，模块或单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性、机械或其它的形式。
作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施方式方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
集成的单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的全部或部分可以以软件产品的形式体 现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)或处理器(processor)执行本申请各个实施方式方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、磁碟或者光盘等各种可以存储程序代码的介质。
工业实用性
本公开实施例提供了一种血管图像检测方法及检测模型训练方法、相关装置、设备，其中，血管图像检测方法包括：获取待测对象的三维血管图像；沿与三维血管图像中血管的延伸方向垂直的方向，从三维血管图像中提取至少一帧二维图像；对二维图像进行检测，得到二维图像的区域检测结果，其中，区域检测结果包括与血管中的斑块对应的第一子区域。根据本公开的实施例的血管图像检测方法，能够能够检测出三维血管图像中属于血管中斑块的像素点，实现像素级别的检测，能够提高血管中斑块的检测精度，且能够有利于为后续量化血管的狭窄程度值提供数据支撑。

Claims (17)

  1. 一种血管图像检测方法，包括：
    获取待测对象的三维血管图像；
    沿与所述三维血管图像中血管的延伸方向垂直的方向，从所述三维血管图像中提取至少一帧二维图像；
    对所述二维图像进行检测，得到所述二维图像的区域检测结果，其中，所述区域检测结果包括与所述血管中的斑块对应的第一子区域。
  2. 根据权利要求1所述的方法，其中，所述区域检测结果还包括与所述血管对应的第一区域；以及
    在所述对所述二维图像进行检测，得到所述二维图像的区域检测结果之后，所述方法还包括：
    基于所述至少一帧二维图像中的所述第一区域和所述第一子区域，得到所述待测对象的所述血管的狭窄程度值。
  3. 根据权利要求2所述的方法，其中，所述基于所述至少一帧二维图像中的所述第一区域和所述第一子区域，得到所述待测对象的所述血管的狭窄程度值，包括：
    利用所述第一子区域的面积和所述第一区域的面积，确定所述狭窄程度值；
    或者，利用所述第一区域在径向方向上最小的第一宽度和所述第一子区域在所述第一区域的径向方向上最小的第二宽度，确定所述狭窄程度值；
    或者，在所述三维血管图像包括多帧二维图像的情况下，利用所述多帧二维图像中所述第一区域的面积与所述第一子区域的面积的差值，确定所述狭窄程度值；
    或者，在所述三维血管图像包括多帧二维图像的情况下，利用所述多帧二维图像中所述第一区域在径向方向上最小的第一宽度与所述第一子区域在所述第一区域的径向方向上最小的第二宽度的差值，确定所述狭窄程度值。
  4. 根据权利要求1至3任一项所述的方法，其中，所述三维血管图像包括多帧所述二维图像，所述对所述二维图像进行检测，得到所述二维图像的区域检测结果，包括：
    分别将每帧所述二维图像作为待测图像，并对所述待测图像和与所述待测图像间隔预设帧数内的二维图像进行检测，得到所述待测图像的区域检测结果。
  5. 根据权利要求4所述的方法，其中，所述区域检测结果还包括与所述血管对应的第一区域；以及
    在所述得到所述待测图像的区域检测结果之后，所述方法还包括：
    按照多帧所述二维图像在所述三维血管图像中的先后顺序，将检测后的多帧所述二维图像进行拼接，得到与所述三维血管图像对应的三维检测图像；
    其中，所述三维检测图像中包含多帧所述二维图像的第一区域拼接得到的第二区域，以及多帧所述二维图像的第一子区域拼接得到的第二子区域；
    利用所述第二区域和所述第二子区域，确定所述待测对象的所述血管中的斑块位置。
  6. 根据权利要求5所述的方法，其中，所述方法还包括：
    利用所述第二子区域的体积和所述第二区域的体积，确定所述待测对象的所述血管的狭窄程度值。
  7. 根据权利要求4至6任一项所述的方法，其中，与所述待测图像间隔预设帧数内的二维图像包括以下至少一者：位于所述待测图像之前的二维图像，位于所述待测图像之后的二维图像。
  8. 根据权利要求1至7任一项所述的方法，其中，所述区域检测结果是利用检测模型对所述二维图像的检测得到的；和/或，
    所述对所述二维图像进行检测，得到所述二维图像的区域检测结果，包括：
    对所述二维图像进行第一检测，确定所述二维图像中是否存在与所述血管对应的第一区域；
    在检测到与所述血管对应的第一区域的情况下，对所述二维图像中的第一区域进行第二检测，得到所述第一区域内的第一子区域。
  9. 根据权利要求8所述的方法，其中，所述对所述二维图像进行第一检测，确定所述二维图像中是否存在与所述血管对应的第一区域，包括：
    利用检测模型的第一检测子网络对所述二维图像进行第一检测，确定所述二维图像中是否存在与所述血管对应的第一区域；
    所述对所述二维图像中的第一区域进行第二检测，得到所述第一区域内的第一子区域，包括：
    利用所述检测模型的第二检测子网络对所述二维图像中的第一区域进行第二检测，得到所述第一区域内的第一子区域。
  10. 根据权利要求9所述的方法，其中，所述第一检测子网络和所述第二检测子网络具有相同的网络结构；
    和/或，所述方法还包括：在未检测到与所述血管对应的第一区域的情况下，对所述二维图像的下一帧图像执行所述第一检测。
  11. 一种检测模型的训练方法，包括：
    获取至少一帧样本二维图像；其中，所述至少一帧样本二维图像是从样本三维血管图像中沿与血管的延伸方向垂直的方向提取得到的，且所述样本二维图像中标注有与所述血管中的斑块对应的第一实际子区域；
    利用所述检测模型对所述样本二维图像进行检测，得到所述样本二维图像中与所述血管中的斑块对应的第一预测子区域；
    利用所述第一实际子区域和所述第一预测子区域的差异，调整所述检测模型的网络参数。
  12. 根据权利要求11所述的方法，其中，所述样本三维血管图像包括多帧样本二维图像；所述利用所述检测模型对所述样本二维图像进行检测，得到所述样本二维图像中与所述血管中的斑块对应的第一预测子区域，包括：
    分别将每帧所述样本二维图像作为样本待测图像，并对所述检测模型对所述样本待测图像和与所述样本待测图像间隔预设帧数内的样本二维图像进行检测，得到所述样本待测图像中的第一预测子区域。
  13. 根据权利要求11或12所述的方法，其中，所述利用所述检测模型对所述样本二维图像进行检测，得到所述样本二维图像中与血管中的斑块对应的第一预测子区域，包括：
    利用所述检测模型的第一检测子网络对所述样本二维图像进行第一检测，确定所述样本二维图像中是否存在与所述血管对应的第一预测区域；
    在检测到与所述血管对应的第一预测区域的情况下，利用所述检测模型的第二检测子网络对所述样本二维图像中的第一预测区域进行第二检测，得到所述第一预测区域内的第一预测子区域。
  14. 一种血管图像检测装置，包括：
    图像获取模块，配置为获取待测对象的三维血管图像；
    图像提取模块，配置为沿与所述三维血管图像中血管的延伸方向垂直的方向，从所述三维血管图像中提取至少一帧二维图像；
    区域检测模块，配置为对所述二维图像进行检测，得到所述二维图像的区域检测结 果，其中，所述区域检测结果包括与所述血管中的斑块对应的第一子区域。
  15. 一种检测模型的训练装置，包括：
    样本获取模块，配置为获取至少一帧样本二维图像；其中，所述至少一帧样本二维图像是从样本三维血管图像中沿与血管的延伸方向垂直的方向提取得到的，且所述样本二维图像中标注有与所述血管中的斑块对应的第一实际子区域；
    区域检测模块，配置为利用所述检测模型对所述样本二维图像进行检测，得到所述样本二维图像中与所述血管中的斑块对应的第一预测子区域；
    参数调整模块，配置为利用所述第一实际子区域和所述第一预测子区域的差异，调整所述检测模型的网络参数。
  16. 一种电子设备，包括相互耦接的存储器和处理器，所述处理器配置为执行所述存储器中存储的程序指令，以实现权利要求1至10任一项所述的血管图像检测方法，或权利要求11至13任一项所述的检测模型的训练方法。
  17. 一种计算机可读存储介质，其上存储有程序指令，所述程序指令被处理器执行时实现权利要求1至10任一项所述的血管图像检测方法，或权利要求11至13任一项所述的检测模型的训练方法。
PCT/CN2021/128160 2021-04-13 2021-11-02 血管图像检测方法及检测模型训练方法、相关装置、设备 WO2022217896A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110396660.6 2021-04-13
CN202110396660.6A CN113096097A (zh) 2021-04-13 2021-04-13 血管图像检测方法及检测模型训练方法、相关装置、设备

Publications (1)

Publication Number Publication Date
WO2022217896A1 true WO2022217896A1 (zh) 2022-10-20

Family

ID=76676981

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/128160 WO2022217896A1 (zh) 2021-04-13 2021-11-02 血管图像检测方法及检测模型训练方法、相关装置、设备

Country Status (2)

Country Link
CN (1) CN113096097A (zh)
WO (1) WO2022217896A1 (zh)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113096097A (zh) * 2021-04-13 2021-07-09 上海商汤智能科技有限公司 血管图像检测方法及检测模型训练方法、相关装置、设备
CN114419047B (zh) * 2022-03-30 2022-07-12 中国科学院自动化研究所 用于确定血管形态特征的方法、装置、设备和存储介质
CN114972242B (zh) * 2022-05-23 2023-04-07 北京医准智能科技有限公司 心肌桥检测模型的训练方法、装置及电子设备

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040228800A1 (en) * 2003-05-15 2004-11-18 Scicotec Gmbh Non-invasive identification of patients at increased risk for myocardial infarction
CN103462696A (zh) * 2013-09-17 2013-12-25 浙江大学 一种集成血管内光相干断层扫描（oct）影像和数字减影(dsa)影像的一体化在线实时处理仪
CN104463830A (zh) * 2013-09-18 2015-03-25 通用电气公司 血管内斑块的侦测系统及方法
CN111862038A (zh) * 2020-07-17 2020-10-30 中国医学科学院阜外医院 一种斑块检测方法、装置、设备及介质
CN113096097A (zh) * 2021-04-13 2021-07-09 上海商汤智能科技有限公司 血管图像检测方法及检测模型训练方法、相关装置、设备

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106447645B (zh) * 2016-04-05 2019-10-15 天津大学 增强ct图像中冠脉钙化检测及量化装置和方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040228800A1 (en) * 2003-05-15 2004-11-18 Scicotec Gmbh Non-invasive identification of patients at increased risk for myocardial infarction
CN103462696A (zh) * 2013-09-17 2013-12-25 浙江大学 一种集成血管内光相干断层扫描（oct）影像和数字减影(dsa)影像的一体化在线实时处理仪
CN104463830A (zh) * 2013-09-18 2015-03-25 通用电气公司 血管内斑块的侦测系统及方法
CN111862038A (zh) * 2020-07-17 2020-10-30 中国医学科学院阜外医院 一种斑块检测方法、装置、设备及介质
CN113096097A (zh) * 2021-04-13 2021-07-09 上海商汤智能科技有限公司 血管图像检测方法及检测模型训练方法、相关装置、设备

Also Published As

Publication number Publication date
CN113096097A (zh) 2021-07-09

Similar Documents

Publication Publication Date Title
WO2022217896A1 (zh) 血管图像检测方法及检测模型训练方法、相关装置、设备
CN111079772B (zh) 图像边缘提取处理方法、装置及存储介质
CN109509186B (zh) 基于大脑ct图像的缺血性脑卒中病灶检测方法及装置
CN109064504B (zh) 图像处理方法、装置和计算机存储介质
CN113077459B (zh) 图像清晰度检测方法、装置、电子设备及存储介质
Harb et al. Improved image magnification algorithm based on Otsu thresholding
CN110378893B (zh) 图像质量评价方法、装置和电子设备
CN108961260B (zh) 图像二值化方法及装置、计算机存储介质
CN110796615A (zh) 一种图像去噪方法、装置以及存储介质
CN106412422A (zh) 对焦方法、装置及终端
CN111507979A (zh) 一种医学影像计算机辅助分析方法
CN111784646A (zh) 图像处理装置及电子设备
CN114782984A (zh) 一种基于tof相机的坐姿识别遮挡判定方法及智能台灯
CN109758170B (zh) X射线成像设备的曝光参数调节方法和装置
CN117224259A (zh) 一种口腔数字印模仪及其扫描头种类识别方法
CN111445411A (zh) 图像去噪方法、装置、计算机设备和存储介质
CN113191965B (zh) 图像降噪方法、设备及计算机存储介质
CN110033466B (zh) 一种基于多灰度级的冠脉拉直图像分割边界确定方法
CN109745073B (zh) 弹性成像位移的二维匹配方法及设备
CN109949243B (zh) 一种钙化伪影消除方法、设备及计算机存储介质
CN115908532A (zh) 线宽识别方法、装置、介质及电子装置
CN112733565A (zh) 二维码粗定位方法、设备及存储介质
CN111192220A (zh) 基于连通域分割的医疗图像降噪方法、装置及存储介质
JP4164166B2 (ja) 照射領域抽出装置、照射領域抽出方法及びコンピュータ読み取り可能な記憶媒体
KR101763376B1 (ko) 신뢰 기반 재귀적 깊이 영상 필터링 방법

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21936763

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21936763

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 13/03/2024)