WO2023124830A1 - 一种基于中心线提取的血管壁影像自动曲面重建方法 - Google Patents

一种基于中心线提取的血管壁影像自动曲面重建方法 Download PDF

Info

Publication number
WO2023124830A1
WO2023124830A1 PCT/CN2022/136953 CN2022136953W WO2023124830A1 WO 2023124830 A1 WO2023124830 A1 WO 2023124830A1 CN 2022136953 W CN2022136953 W CN 2022136953W WO 2023124830 A1 WO2023124830 A1 WO 2023124830A1
Authority
WO
WIPO (PCT)
Prior art keywords
blood vessel
image
centerline
detected
blood
Prior art date
Application number
PCT/CN2022/136953
Other languages
English (en)
French (fr)
Inventor
张娜
刘新
郑海荣
申帅
梁栋
胡战利
李烨
邹超
Original Assignee
深圳先进技术研究院
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 深圳先进技术研究院 filed Critical 深圳先进技术研究院
Publication of WO2023124830A1 publication Critical patent/WO2023124830A1/zh

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/0033—Features or image-related aspects of imaging apparatus classified in A61B5/00, e.g. for MRI, optical tomography or impedance tomography apparatus; arrangements of imaging apparatus in a room
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/05—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves 
    • A61B5/055—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves  involving electronic [EMR] or nuclear [NMR] magnetic resonance, e.g. magnetic resonance imaging
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10028—Range image; Depth image; 3D point clouds
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10072—Tomographic images
    • G06T2207/10088—Magnetic resonance imaging [MRI]
    • 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 field of graphic reconstruction, in particular to a method for automatic surface reconstruction of blood vessel wall images based on centerline extraction.
  • Stroke is the disease with the highest mortality and disability rate in my country, and has become a serious burden on my country's medical expenditure and social families. Stroke includes hemorrhagic stroke and ischemic stroke. In my country, ischemic stroke is the main type of stroke, accounting for 79% of all stroke cases, and there is an increasing trend. Studies have shown that thrombosis caused by atherosclerotic plaque rupture is the main pathogenesis of ischemic stroke, of which 48% of culprit plaques originate from intracranial arteries, 30% originate from carotid arteries, and the remaining 22% mainly originate from the heart. and thoracic aorta. Therefore, timely detection of stroke-related vascular beds, including vulnerable plaques or other vessel wall lesions in intracranial arteries, carotid arteries, and thoracic aorta, is the key to early prevention and precise treatment of ischemic stroke.
  • 3D reconstruction of the relevant vessel walls is required. Due to the curvature of the arteries, especially the repeated winding of the intracranial arteries, the entire vessel wall cannot be displayed in the same plane. Therefore, a good method is to extract the blood vessel centerline coordinates from the MR image, use the centerline as a map, identify the position of the cross-sectional image with blood vessels, and obtain a series of continuous slices with blood vessels in the center for the blood vessel Wall division and other tasks.
  • Bright blood MR images are usually used to extract the centerline of blood vessels for vessel reconstruction.
  • Bright blood MR images including 3D-TOF MRA which makes the blood flow bright and high signal, is mainly used to screen for intracranial vascular stenosis.
  • the high blood flow signal in the bright blood MR image makes it difficult to distinguish the outer wall of the blood vessel from other tissues, it is usually necessary to use the extracted blood vessel centerline to reconstruct the corresponding blood vessel wall in the black blood MR image.
  • Black blood MR image refers to the MR image obtained by using the static characteristics of blood flow and blood vessel wall to suppress the flowing blood signal, making the blood flow low signal, the vessel wall and plaque high signal, and enhancing the contrast of the inner wall of the blood vessel.
  • the blood flow signal is suppressed in the black blood MR image, it is not easy to directly extract the centerline of the blood vessel.
  • the inventor believes that since the imaging time of a single MRI sequence is usually about a few minutes, and the resolution at the molecular level can be achieved, it is susceptible to errors caused by various influences, such as vascular pulsation, muscle movement, human body movement, and sudden changes in blood flow etc., and will often cause image artifacts in a single imaging.
  • the above-mentioned black blood MR and bright blood MR are imaged twice independently, and the blood vessel centerline obtained by the above technology is prone to misregistration problems when applied to black blood images, especially in intracranial arteries, especially It is especially serious on intracranial perforating arteries.
  • the present application provides a method for automatic surface reconstruction of vessel wall images based on centerline extraction.
  • the present application provides a blood vessel centerline extraction method for three-dimensional reconstruction of blood vessel wall in black blood MR images, which adopts the following technical scheme:
  • a blood vessel centerline extraction method for three-dimensional reconstruction of blood vessel walls in black blood MR images comprising the following steps:
  • the image to be detected is a black blood MR image
  • the first blood vessel feature is obtained from the image to be detected, wherein the first target detection model is trained to extract features from the two-dimensional image, and the first blood vessel feature is used to roughly correspond to the center of the blood vessel The position of the line on the image to be detected;
  • the image to be detected superimposed with the first blood vessel feature is sent to the three-dimensional image segmentation model to obtain the second blood vessel feature, wherein the three-dimensional image segmentation model is used to extract three-dimensional feature information based on the three-dimensional convolution kernel, and the second blood vessel feature is used corresponds to the three-dimensional spatial position of the centerline of the blood vessel on the image to be detected;
  • the blood vessel centerline coordinates are obtained based on the second target detection model and the second blood vessel feature.
  • the black blood MR image is a three-dimensional image, equivalent to being composed of multiple consecutive two-dimensional images.
  • the target detection model samples the image to be detected into multiple two-dimensional images and processes them separately.
  • the target detection model removes the background from the image to be detected to obtain the vessel area initially, and calculates the centerline of the vessel based on the vessel area as the first vessel feature. . Since the target detection model can only extract the features of two-dimensional images, the detection results show that there are many discontinuous blood vessels and isolated data. Therefore, this step achieves a relatively rough centerline of the vessel.
  • the target detection model can only extract the features of two-dimensional images, it generates few intermediate parameters, and has the advantages of low computing cost and fast processing speed.
  • the first blood vessel feature is added to the image to be detected to form two channel data, so that the 3D image segmentation model can perform feature recognition to extract the second blood vessel feature.
  • the 3D image segmentation model uses 3D convolution, which can effectively use 3D features and ensure the continuity of blood vessel prediction results.
  • three-dimensional convolution requires high computing power, so it is used to optimize the rough segmentation results output by the target detection model.
  • the vascular wall outline output by the 3D image segmentation model is fine enough, it cannot provide enough information (not including coordinates) for making blood vessel slices. Therefore, it is necessary to connect another target detection model to output the coordinates of the blood vessel slice.
  • three models are used to continuously process black blood MR images, so as to accurately obtain the continuous coordinates of the vessel centerline in space, which can be used for later reconstruction of vessel wall images using black blood MR images, without the need for different sequences
  • the registration of MR images avoids the resulting errors; at the same time, it can improve the performance of subsequent three-dimensional reconstruction of vessel walls.
  • the first blood vessel feature is N rectangular frames determined in six dimensions, where N is the Nth target appearing in the image to be detected, X is the x-axis coordinate of the predicted target center, and Y is The y-axis coordinate of the predicted target center, W is the width of the predicted target, H is the height of the predicted target, and C is the category of the recognized target.
  • the target detection model is a Yolo model, an SSD model, a Mask-R-CNN model or a Fast-R-CNN model.
  • the Yolo model, SSD model, Mask-R-CNN model or Fast-R-CNN model are target detection models for 2D images, used for target recognition, and they are characterized by lower computing costs and faster processing speed.
  • the 3D image segmentation model is a V-net model.
  • the V-net model uses three-dimensional convolution, which can effectively use three-dimensional features and ensure the continuity of blood vessel prediction results.
  • three-dimensional convolution requires high computing power, so it is used to optimize the rough segmentation results output by Yolo.
  • the present application provides a method for automatic surface reconstruction of vessel wall images based on centerline extraction, which adopts the following technical solution:
  • a method for automatic surface reconstruction of vessel wall images based on centerline extraction comprising the following steps:
  • the slices are superimposed along the z-axis to obtain the result of surface reconstruction.
  • the blood vessel centerline output by the deep learning model is composed of a series of (three-dimensional) points, so it needs to be fitted to a straight line to represent the direction of the blood vessel. Due to the tortuous trend of blood vessels, it is difficult to show the whole picture on a single cut plane, and even multiple end faces may be generated.
  • the blood vessels can be straightened to reflect the characteristics of the whole blood vessels. Therefore, using the center point of the blood vessel and the center line of the blood vessel to obtain slices on the normal plane, and then superimposing the slices along the z-axis, the purpose of surface reconstruction and straightening of the blood vessel is completed.
  • the method for fitting the coordinates of each point of the blood vessel centerline is B-spline fitting.
  • the sampling range is selected based on the vessel type.
  • an electronic device provided by this application adopts the following technical solution:
  • a blood vessel centerline extraction system for three-dimensional reconstruction of blood vessel walls in black blood MR images including:
  • the input module is used to obtain the image to be detected; wherein, the image to be detected is a black blood MR image;
  • the first detection module is used to obtain the first blood vessel feature from the image to be detected based on the Yolo model, wherein the first blood vessel feature is used to roughly correspond to the position of the center line of the blood vessel on the image to be detected;
  • a combination module used to superimpose the first blood vessel feature to the image to be detected and form two channel data
  • the second detection module is used to send the image to be detected superimposed with the first blood vessel feature into the V-net model to obtain the second blood vessel feature, wherein the second blood vessel feature is used to finely correspond to the center line of the blood vessel in the image to be detected position on
  • the third detection module is used to obtain the blood vessel centerline coordinates based on the Yolo network and the second blood vessel feature.
  • an electronic device provided by this application adopts the following technical solution:
  • An automatic surface reconstruction system for vessel wall images based on centerline extraction including:
  • the above-mentioned blood vessel centerline extraction system is used to obtain the coordinates of each point of the blood vessel centerline;
  • Fitting module used for fitting the coordinates of each point of the blood vessel centerline
  • the modeling module is used to calculate the normal plane passing through each blood vessel center point based on the fitting result, and establish a space coordinate system based on each normal plane, wherein the z-axis is the normal direction of the normal plane;
  • a sampling module configured to sample the vessel wall image based on a normal plane to obtain slices
  • the reconstruction module is configured to superimpose the slices along the z-axis based on the order of the points on the centerline of the blood vessel to obtain a surface reconstruction result.
  • an electronic device provided by this application adopts the following technical solution:
  • An electronic device comprising:
  • processors one or more processors
  • one or more application programs wherein the one or more application programs are stored in the memory and configured to be executed by the one or more processors, the one or more program programs are configured to:
  • a computer-readable storage medium provided by the present application adopts the following technical solution:
  • a computer-readable storage medium stores a computer program capable of being loaded by a processor and executing the above-mentioned method.
  • the storage medium stores at least one instruction, at least one section of program, code set or instruction set, and the at least one instruction, the at least one section of program, the code set or instruction set are loaded and executed by the processor to implement:
  • FIG. 1 is used to illustrate the steps of a method for automatic surface reconstruction of a vessel wall image based on centerline extraction in an embodiment of the present application.
  • FIG. 2 is used to illustrate the steps of a blood vessel centerline extraction method for three-dimensional reconstruction of blood vessel walls from black blood MR images in an embodiment of the present application.
  • FIG. 3 is used to illustrate a black blood MR image as an image to be detected.
  • FIG. 4 is used to illustrate the first vessel features obtained from the image to be detected by using the first object detection model, wherein the box at the intersection is used to indicate the detected position of the left common carotid artery-internal carotid artery bifurcation.
  • FIG. 5 is used to illustrate the first vessel features obtained from the image to be detected by using the first target detection model, wherein the box at the intersection is used to indicate the detected position of the left common carotid artery-external carotid artery bifurcation.
  • FIG. 6 is used to show that the first blood vessel feature is filled in the image to be detected, and the square is used to indicate the left common carotid artery-internal carotid artery bifurcation.
  • FIG. 7 is used to show that the first blood vessel feature is filled in the image to be detected, and the squares are used to indicate the left common carotid artery-external carotid artery bifurcation.
  • FIG. 8 is used to illustrate an image generated by superimposing the first blood vessel feature on the image to be detected.
  • FIG. 9 is used to show the output image after the V-net model is trained on the superimposed image.
  • each step in this embodiment is for convenience of description only, and do not represent a limitation on the execution order of each step.
  • the execution order of each step can be adjusted according to needs, or performed at the same time. These adjustments or replacements belong to the present invention scope of protection.
  • the embodiment of the present application discloses a method for automatic surface reconstruction of blood vessel wall images based on centerline extraction, referring to FIG. 1 , including the following steps:
  • the fitting method may be a method such as B-spline fitting, but any method capable of fitting multiple points into a continuous curve may be used.
  • Magnetic resonance imaging resonance imaging is the use of nuclear magnetic resonance (nuclear magnetic Resonance, NMR) principle, according to the different attenuation of the released energy in different structural environments inside the material, by detecting the emitted electromagnetic waves with an external gradient magnetic field, the position and type of the atomic nucleus that constitutes the object can be known. Mapped as an image of the structure inside the object.
  • the imaging time of a single MRI sequence is usually about a few minutes, and molecular-level resolution can be achieved. Therefore, for images obtained by different MRIs, such as black blood MR images and bright blood MR images of the same subject, bright blood is prone to occur.
  • the vascular centerline coordinates extracted from the MR image do not match the blood vessel pattern in the black blood MR image. Therefore, in step S1 , a blood vessel centerline extraction method for three-dimensional reconstruction of blood vessel walls from black blood MR images can be used to directly process the black blood MR images to obtain blood vessel centerlines.
  • the blood vessel centerline extraction method for three-dimensional reconstruction of blood vessel wall in black blood MR images includes the following steps:
  • the image to be detected is an image obtained by vascular wall magnetic resonance imaging (VW MRI).
  • VW MRI vascular wall magnetic resonance imaging
  • the image to be detected is a two-dimensional RGB image (three-dimensional matrix), that is, the Channel is an array, and the size of the dimension is determined to be 3.
  • the range of the three-dimensional data corresponding to the Channel is usually 0-65535, which represents the corresponding speed-up. Signal intensity during resonance imaging.
  • the first target detection model can be Yolo model, SSD model, Mask-R-CNN model or Fast-R-CNN model or other models, and these models are all target detection models for 2D images , used for target recognition, they are characterized by lower computational cost and faster processing speed.
  • the first target detection model is the Yolo model.
  • the first blood vessel feature is output, which is the result of the Yolo network.
  • the result is the coordinates of a rectangle tangent to the blood vessel, including coordinates and width and height information.
  • the first blood vessel feature is N rectangular boxes determined in six dimensions, and they frame N objects identified by the neural network.
  • N is the Nth target appearing in the image to be detected
  • X is the x-axis coordinate of the predicted target center
  • Y is the y-axis coordinate of the predicted target center
  • W is the width of the predicted target
  • H is The height of the predicted target
  • C is the category of the recognized target. Note that there is only one type of target in this method.
  • the Yolo model can only extract the features of two-dimensional images, the detection results show that there are many discontinuous blood vessels and isolated data, so there will be N targets. Therefore, a relatively rough vessel centerline is obtained in this step.
  • the first target detection model obtains the first vessel features from the image to be detected, which are respectively the left common carotid artery-internal carotid artery bifurcation corresponding to Figure 4, and Figure 5 The corresponding left common carotid artery-external carotid artery bifurcation.
  • the conversion method is to assign a value of 1 to the rectangle corresponding to the Yolo detection result, and assign a value of 0 to the rest.
  • the five dimensions are:, under the premise of ensuring that the dimensions of Batch, Z, Y, and X are the same, only the second dimension of the matrix needs to be spliced.
  • Figure 6 the left common carotid artery-internal carotid artery bifurcation in the original image is filled based on the first blood vessel feature (coordinate), and Figure 7 is based on the first blood vessel feature (coordinate ) is filled in the left common carotid artery-external carotid artery bifurcation in the original image.
  • the 3D image segmentation model adopts the V-net model.
  • V-net uses three-dimensional convolution, which can effectively use three-dimensional features to ensure the continuity of blood vessel prediction results.
  • three-dimensional convolution requires high computing power, so it is used to optimize the rough segmentation results output by Yolo.
  • the output of V-net has the same dimensions as the original data, which is used as an intermediate result.
  • the generated first blood vessel feature is superimposed on the original image to obtain the image shown in Figure 8, and then input into the V-net model for training to obtain the image shown in Figure 9.
  • the second target detection model can be Yolo model, SSD model, Mask-R-CNN model or Fast-R-CNN model or other models, and these models are all used for 2D images Object detection models, used for object recognition, they are characterized by low computational cost and fast processing speed. As an example, in this embodiment, the second target detection model also uses the Yolo model.
  • the normal plane passing through the center point of the blood vessel is obtained.
  • the center point of the blood vessel is taken as the origin
  • the direction perpendicular to the y-axis in the original data is defined as the x-axis of the normal plane
  • the direction perpendicular to the x-axis is defined as the y-axis of the normal plane
  • the direction perpendicular to the normal plane is defined as z-axis.
  • sampling range As small as possible. For example, under the imaging condition of 0.33mm/px, you can choose a sampling range of 64x64 centered on the origin. A slice is obtained according to the sampling range (variable) on the normal plane of each blood vessel center point, which is called a slice.
  • each slice is superimposed along the z-axis to obtain a result of surface reconstruction.
  • the vessel centerline output by the deep learning model is composed of a series of (three-dimensional) points, which are fitted to a straight line to represent the direction of the vessel. Due to the tortuous trend of blood vessels, it is difficult to show the whole picture on a single cut plane, and even multiple end faces may be generated. After surface reconstruction, the blood vessels can be straightened to reflect the characteristics of the whole blood vessels. Therefore, using the center point of the blood vessel and the center line of the blood vessel to obtain slices on the normal plane, and then superimposing the slices along the z-axis, the purpose of surface reconstruction and straightening of the blood vessel is completed.
  • the embodiment of the present application also discloses an electronic device, including a memory and a processor, the memory stores a method for extracting the centerline of a blood vessel that can be loaded by the processor and executed as described above for three-dimensional reconstruction of blood vessel walls in black blood MR images, and /or a computer program based on the method of automatic surface reconstruction of vessel wall images based on centerline extraction.
  • the execution subject of the method in this embodiment may be a control device, which is set on an electronic device.
  • the current device may be a mobile phone, a tablet computer, a notebook computer and other electronic devices with a WIFI function.
  • the execution subject of the method in this embodiment is also It may directly be a CPU (central processing unit, central processing unit) of an electronic device.
  • the embodiment of the present application also discloses a computer-readable storage medium, which stores the method for extracting the centerline of blood vessels for three-dimensional reconstruction of blood vessel walls in black blood MR images, and/or the extraction method based on the centerline, which can be loaded and executed by the processor.
  • a computer program for an automatic curved surface reconstruction method for vessel wall images can be implemented by means of software plus a necessary general-purpose hardware platform, and of course also by hardware, but in many cases the former is better implementation.
  • the technical solution of the present application can be embodied in the form of a software product in essence or in other words, the part that contributes to the prior art, and the computer software product is stored in one of the above storage media (such as ROM/RAM, magnetic CD, CD), including several instructions to make a device (which can be a mobile phone, computer, server, controlled terminal, or network device, etc.) execute the method of each embodiment of the present application.
  • a software product in essence or in other words, the part that contributes to the prior art
  • the computer software product is stored in one of the above storage media (such as ROM/RAM, magnetic CD, CD), including several instructions to make a device (which can be a mobile phone, computer, server, controlled terminal, or network device, etc.) execute the method of each embodiment of the present application.

Landscapes

  • Health & Medical Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Surgery (AREA)
  • Veterinary Medicine (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Medical Informatics (AREA)
  • Molecular Biology (AREA)
  • Pathology (AREA)
  • Animal Behavior & Ethology (AREA)
  • General Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Biomedical Technology (AREA)
  • Geometry (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Biophysics (AREA)
  • Radiology & Medical Imaging (AREA)
  • High Energy & Nuclear Physics (AREA)
  • Computer Graphics (AREA)
  • Software Systems (AREA)
  • Magnetic Resonance Imaging Apparatus (AREA)

Abstract

本申请涉及一种基于中心线提取的血管壁影像自动曲面重建方法，其包括以下步骤：利用血管中心线提取方法获取血管中心线的各点坐标并拟合；基于拟合结果计算经过各血管中心点的法平面，并基于各法平面分别建立空间坐标系，其中z轴为与法平面的法线方向；基于法平面对血管壁影像进行采样以获得切片；基于血管中心线各点的顺序将各切片沿z轴叠加，得到曲面重建的结果。本申请具有血管壁重建效果好，经过曲面重建可以将血管拉直以反映血管整体特征的优点。

Description

一种基于中心线提取的血管壁影像自动曲面重建方法 技术领域
本申请涉及图形重建的领域，尤其是涉及一种基于中心线提取的血管壁影像自动曲面重建方法。
 背景技术
脑卒中是我国死亡率和致残率最高的疾病，已成为我国医疗支出和社会家庭的严重负担。脑卒中包括出血性脑卒中和缺血性脑卒中两种，我国脑卒中以缺血性脑卒中为主，占全部脑卒中病例的79%，而且有增加趋势。研究表明，动脉粥样硬化斑块破裂引发血栓形成是缺血性脑卒中的主要发病机制，其中罪犯斑块48%来源于颅内动脉，30%来源于颈动脉，其余22%主要来源于心脏和胸主动脉。因此，及时发现脑卒中相关血管床，包括颅内动脉、颈动脉和胸主动脉的易损斑块或其他管壁病变是缺血性脑卒中早期预防和精准治疗的关键。
为了进行脑卒中相关血管床的自动化分析任务，需要对相关血管壁进行三维重建。由于动脉形态弯曲，尤其是颅内动脉的走行反复卷绕，不能在同一个平面内显示血管壁的全程。因此，一个好的方法是从MR影像中提取血管中心线坐标，利用该中心线为地图，对存在血管的横断面影像的进行位置标识，并获取一系列血管位于中心的连续切片，用于血管壁的分割等任务。
在相关技术中，通常采用亮血MR影像来对血管中心线进行提取以用于血管重建。亮血MR影像包括3D-TOF MRA，使血流呈明亮高信号，主要用于筛查颅内血管狭窄。由于亮血MR影像中过高的血流信号使血管外壁与其他组织难以区分，通常还要利用提取的血管中心线对相应的黑血MR影像的血管壁进行重建。黑血MR影像指利用血液流动和血管壁静止的特性，抑制流动血液信号，使血流呈低信号，管壁及斑块呈较高信号，增强血管内壁的对比度，所得到的MR影像。但由于黑血MR影像中血流信号被抑制，因此不容易直接提取血管中心线。
发明人认为，由于单个磁共振序列的成像时间通常在几分钟左右，并且可以达到分子级的分辨率，因此容易受到多方面的影响产生误差，比如血管搏动、肌肉运动、人体运动、血液流量突变等，并且在单次成像中都会经常引起影像伪影的产生。而上述的黑血MR和亮血MR是两次独立进行成像的，通过上述技术取得的血管中心线在应用到黑血影像时很容易出现不能配准的问题，尤其在颅内动脉，特别是颅内穿支动脉上尤其严重。
 技术问题
为了提高血管壁重建效果，本申请提供一种基于中心线提取的血管壁影像自动曲面重建方法。
 技术解决方案
第一方面，本申请提供的一种用于黑血MR影像血管壁三维重建的血管中心线提取方法，采用如下的技术方案：
一种用于黑血MR影像血管壁三维重建的血管中心线提取方法，包括以下步骤：
获取待检测影像；其中，所述的待检测影像为黑血MR影像；
基于第一目标检测模型从将待检测影像中获取第一血管特征，其中，所述第一目标检测模型被训练用于从二维影像中提取特征，第一血管特征用于粗略对应于血管中心线在待检测影像上的位置；
叠加第一血管特征至待检测影像并形成两个通道数据；
将叠加有第一血管特征的待检测影像送入三维影像分割模型，以获得第二血管特征，其中，所述三维影像分割模型用于基于三维卷积核提取三维特征信息，第二血管特征用于精细对应于血管中心线在待检测影像上的三维空间位置；
基于第二目标检测模型和第二血管特征获得血管中心线坐标。
通过采用上述技术方案，黑血MR影像为三维影像，相当于由连续多张二维影像组成。目标检测模型将待检测影像抽样成多张二维影像并分别进行处理，通过目标检测模型从待检测影像中进行去除背景以初步获取血管区域，并基于血管区域计算出血管中心线并作为第一血管特征。由于目标检测模型只能提取二维影像的特征，检测的结果显示有许多不连续的血管和孤立的数据。因此，这一步取得了一个相对粗略的血管中心线。另外，由于目标检测模型只能提取二维影像的特征，产生的中间参数少，具有运算成本低和处理速度快的优点。
再将第一血管特征至待检测影像并形成两个通道数据，以便于三维影像分割模型进行特征识别以提取出第二血管特征。由于经过目标检测模型的处理，相当于提供先验条件，三维影像分割模型使用的是三维卷积，可以有效地利用三维特征，保证血管预测结果的连续性。但三维卷积需要的算力较高，因此用来优化目标检测模型输出的粗分割结果。
最后，由于三维影像分割模型输出的血管壁轮廓已经足够精细了，但无法为制作血管切片提供足够的信息（不包含坐标）。因此需要再接一个目标检测模型输出血管切片的坐标。
综上，通过三个模型对黑血MR影像进行连续处理，从而精确地获取血管中心线在空间上的连续坐标，以用于后期利用黑血MR影像进行血管壁影像重建，不需要进行不同序列MR影像的配准，避免了因此带来的误差；同时，可以提升后续血管壁三维重建工作的性能。
可选的，所述第一血管特征为六个维度确定的N个矩形框，其中，N为待检测影像中出现的第N个目标，X为预测出的目标中心的x轴坐标，Y为预测出的目标中心的y轴坐标，W为预测出的目标的宽度，H为预测出的目标的高度，C为识别到的目标的类别。
可选的，所述目标检测模型为Yolo模型、SSD模型、Mask-R-CNN模型或Fast-R-CNN模型。
通过采用上述技术方案，Yolo模型、SSD模型、Mask-R-CNN模型或Fast-R-CNN模型均是用于2D影像的目标检测模型，用于目标识别，它们的特点是较低的运算成本和较快的处理速度。
可选的，所述三维影像分割模型为V-net模型。
通过采用上述技术方案，V-net模型使用的是三维卷积，可以有效地利用三维特征，保证血管预测结果的连续性。但三维卷积需要的算力较高，因此用来优化Yolo输出的粗分割结果。
第二方面，本申请提供的一种基于中心线提取的血管壁影像自动曲面重建方法，采用如下的技术方案：
一种基于中心线提取的血管壁影像自动曲面重建方法，包括以下步骤：
基于上述的血管中心线提取方法获取血管中心点的坐标并拟合；
基于拟合结果计算经过各血管中心点的法平面，并基于各法平面分别建立空间坐标系，其中z轴为与法平面的法线方向；
基于法平面对血管壁影像进行采样以获得切片；
基于血管中心线各点的顺序将各切片沿z轴叠加，得到曲面重建的结果。
通过采用上述技术方案，由深度学习模型输出的血管中心线是由一系列（三维的）点组成的，因此需要对其进行拟合为直线以用于表征血管的走向。由于血管得走势迂回，在单个切面上难以显示其全貌，甚至会产生多个端面，经过曲面重建可以将血管拉直，可以反映血管整体的特征。因此，利用血管中心点和血管中心线获取法平面上的切片，再将切片沿z轴叠加，即完成曲面重建和将血管拉直的目的。
可选的，所述的拟合血管中心线的各点坐标的方法为B样条拟合。
可选的，在基于法平面对血管壁影像进行采样以获得切片之前，还包括以下步骤：
基于血管类型选取采样范围。
通过采用上述技术方案，在不丢失血管壁所在像素的前提下，选定一个尽量小的采样范围，能有有效减少数据处理量并使得曲面重建的结果更为直观。
第三方面，本申请提供的一种电子设备，采用如下的技术方案：
一种用于黑血MR影像血管壁三维重建的血管中心线提取系统，包括：
输入模块，用于获取待检测影像；其中，所述的待检测影像为黑血MR影像；
第一检测模块，用于基于Yolo模型从将待检测影像中获取第一血管特征，其中，第一血管特征用于粗略对应于血管中心线在待检测影像上的位置；
组合模块，用于叠加第一血管特征至待检测影像并形成两个通道数据；
第二检测模块，用于将叠加有第一血管特征的待检测影像送入V-net模型，以获得第二血管特征，其中，第二血管特征用于精细对应于血管中心线在待检测影像上的位置；
第三检测模块，用于基于Yolo网络和第二血管特征获得血管中心线坐标。
第四方面，本申请提供的一种电子设备，采用如下的技术方案：
一种基于中心线提取的血管壁影像自动曲面重建系统，包括：
如上述的血管中心线提取系统，用于获取血管中心线的各点坐标；
拟合模块，用于拟合血管中心线的各点坐标；
建模模块，用于基于拟合结果计算经过各血管中心点的法平面，并基于各法平面分别建立空间坐标系，其中，z轴为与法平面的法线方向；
采样模块，用于基于法平面对血管壁影像进行采样以获得切片；
重建模块，用于基于血管中心线各点的顺序将各切片沿z轴叠加，得到曲面重建的结果。
第五方面，本申请提供的一种电子设备，采用如下的技术方案：
一种电子设备，其包括：
一个或多个处理器；
存储器；
一个或多个应用程序，其中所述一个或多个应用程序被存储在所述存储器中并被配置为由所述一个或多个处理器执行，所述一个或多个程序配置用于：
执行上述的血管中心线提取方法；
和/或，执行上述的血管壁影像自动曲面重建方法。
第六方面，本申请提供的一种计算机可读存储介质，采用如下的技术方案：
一种计算机可读存储介质，存储有能够被处理器加载并执行如上的上述方法的计算机程序。
所述存储介质存储有至少一条指令、至少一段程序、代码集或指令集，所述至少一条指令、所述至少一段程序、所述代码集或指令集由所述处理器加载并执行以实现：
如上述的血管中心线提取方法；
和/或，上述的血管壁影像自动曲面重建方法。
 附图说明
图1用于示出本申请实施例中一种基于中心线提取的血管壁影像自动曲面重建方法的步骤。
图2用于示出本申请实施例中一种黑血MR影像血管壁三维重建的血管中心线提取方法的步骤。
图3用于示出一种作为待检测影像的黑血MR影像。
图4用于示出利用第一目标检测模型从待检测影像获取的第一血管特征，其中十字交叉处的方框用于示意检测到的左侧颈总动脉-颈内动脉分叉的位置。
图5用于示出利用第一目标检测模型从待检测影像获取的第一血管特征，其中十字交叉处的方框用于示意检测到的左侧颈总动脉-颈外动脉分叉的位置。
图6用于示出第一血管特征填充在待检测影像，其中方块用于示意左侧颈总动脉-颈内动脉分叉处。
图7用于示出第一血管特征填充在待检测影像，其中方块用于示意左侧颈总动脉-颈外动脉分叉。
图8用于示出第一血管特征叠加在待检测影像所生成影像。
图9用于示出V-net模型对叠加影像训练后输出的影像。
 本发明的实施方式
以下结合附图，对本申请作进一步详细说明。应当理解，此处所描述的具体实施例仅仅用以解释本申请，并不用于限定本申请。
本实施例中各步骤的标号仅为方便说明，不代表对各步骤执行顺序的限定，在实际应用时，可以根据需要各步骤执行顺序进行调整，或同时进行，这些调整或者替换均属于本发明的保护范围。
在以下描述中，为了解释的目的，阐述了很多具体细节，以便提供对发明构思的彻底理解。作为本说明书的一部分，本公开的附图中的一些附图以框图形式表示结构和设备，以避免使所公开的原理复杂难懂。为了清晰起见，实际具体实施的并非所有特征都有必要进行描述。此外，本公开中所使用的语言已主要被选择用于可读性和指导性目的，并且可能没有被选择为划定或限定本发明的主题，从而诉诸于所必需的权利要求以确定此类发明主题。在本公开中对“一个具体实施”或“具体实施”的提及意指结合该具体实施所述的特定特征、结构或特性被包括在至少一个具体实施中，并且对“一个具体实施”或“具体实施”的多个提及不应被理解为必然地全部是指同一具体实施。
本申请实施例公开一种基于中心线提取的血管壁影像自动曲面重建方法，参照图1，包括以下步骤：
S1.获取血管中心点的坐标并拟合。其中，拟合方法可以为B样条拟合等方法，但凡能够将多点拟合为连续曲线的方法均可。
磁共振成像（Magnetic resonance imaging，MRI），是利用核磁共振（nuclear magnetic resonance，NMR）原理，依据所释放的能量在物质内部不同结构环境中不同的衰减，通过外加梯度磁场检测所发射出的电磁波，即可得知构成这一物体原子核的位置和种类，据此可以绘制成物体内部的结构影像。单个磁共振序列的成像时间通常在几分钟左右，并且可以达到分子级的分辨率，因此对于不同的核磁共振获得的影像，比如同一对象的黑血MR影像和亮血MR影像，容易发生亮血MR影像提取出的血管中心线坐标与黑血MR影像内的血管图案不匹配的情况。因此，在步骤S1中可以利用一种用于黑血MR影像血管壁三维重建的血管中心线提取方法，直接对黑血MR影像进行处理以获得血管中心线。
具体的，参照图2，该用于黑血MR影像血管壁三维重建的血管中心线提取方法包括以下步骤：
S11.获取待检测影像；其中，所述的待检测影像为黑血MR影像。
参照图3，该待检测影像是血管壁磁共振成像（VW MRI）得到的影像。待检测影像为二维RGB影像 (三维矩阵)，即，其中Channel为数组，维度的大小确定为3，具体的，Channel所对应的三维数据的范围通常是0-65535，代表了相应提速实在磁共振成像时的信号强度。
S12.基于第一目标检测模型从将待检测影像中获取第一血管特征，其中，所述第一目标检测模型被训练用于从二维影像中提取特征，第一血管特征用于粗略对应于血管中心线在待检测影像上的位置。
在不同的实施例中，第一目标检测模型可以为Yolo模型、SSD模型、Mask-R-CNN模型或Fast-R-CNN模型或其它模型，该类模型均是用于2D影像的目标检测模型，用于目标识别，它们的特点是较低的运算成本和较快的处理速度。作为示例的，在本实施例中，第一目标检测模型选用Yolo模型。
经过Yolo网络的处理，输出第一血管特征，也就是Yolo网络的结果。该结果是与血管相切的矩形的坐标，包括坐标和宽度、高度信息。具体的，第一血管特征为六个维度确定的N个矩形框，它们框出的是神经网络识别到的N个目标。其中，N为待检测影像中出现的第N个目标，X为预测出的目标中心的x轴坐标，Y为预测出的目标中心的y轴坐标，W为预测出的目标的宽度，H为预测出的目标的高度，C为识别到的目标的类别。需要注意的是，在本方法中只有一类目标。
由于Yolo模型只能提取二维影像的特征，检测的结果显示有许多不连续的血管和孤立的数据，因此就会有N个目标。因此，在该步骤中取得了一个相对粗略的血管中心线。
举个例子，参照图4和图5，第一目标检测模型从将待检测影像中获取第一血管特征，分别为图4所对应的左侧颈总动脉-颈内动脉分叉，以及图5所对应的左侧颈总动脉-颈外动脉分叉。
S13.叠加第一血管特征至待检测影像并形成两个通道数据。
将Yolo的输出结果转换成与原数据相同的格式在叠加。转换的方法是将Yolo检测结果对应的矩形赋值为1，其余部分赋值为0。
在进行3D卷积时，在网络内部通常是使用五维数据进行运算和推理。五个维度分别为：，在保证Batch, Z, Y, X四个维度尺寸相同的前提下，只需要对矩阵的第二个维度拼接即可。
举个例子，参考图6和图7，图6中为基于第一血管特征（坐标）填充在原图的左侧颈总动脉-颈内动脉分叉处，图7为基于第一血管特征（坐标）填充在原图的左侧颈总动脉-颈外动脉分叉处。
S14.将叠加有第一血管特征的待检测影像送入三维影像分割模型，以获得第二血管特征，其中，所述三维影像分割模型用于基于三维卷积核提取三维特征信息，第二血管特征用于精细对应于血管中心线在待检测影像上的三维空间位置。
在本实施例中，三维影像分割模型采用V-net模型。V-net使用的是三维卷积，可以有效地利用三维特征，保证血管预测结果的连续性。但三维卷积需要的算力较高，因此用来优化Yolo输出的粗分割结果。V-net的输出与原始数据的维度一样，将其作为中间结果。
举个例子，对生成的第一血管特征（影像）与原始影像叠加，得到如图8所示的影像，再输入V-net模型进行训练，得到如图9所示的影像。
S15.基于第二目标检测模型和第二血管特征获得血管中心线坐标。
在该步骤中，将中间结果输入第二目标检测模型，或者将中间结果和原始数据叠加输入网络都可以得到准确的血管中心线坐标。同样的，在不同的实施例中，第二目标检测模型可以为Yolo模型、SSD模型、Mask-R-CNN模型或Fast-R-CNN模型或其它模型，该类模型均是用于2D影像的目标检测模型，用于目标识别，它们的特点是较低的运算成本和较快的处理速度。作为示例的，在本实施例中，第二目标检测模型也选用Yolo模型。
这里之所以需要第二个Yolo网络，是因为V-net输出的血管壁轮廓已经足够精细了，但无法为制作血管切片提供足够的信息（不包含坐标）。因此需要再接一个Yolo模型用于输出血管切片的坐标，即血管中心点坐标。
S2.基于拟合结果计算经过各血管中心点的法平面，并基于各法平面分别建立空间坐标系，其中z轴为与法平面的法线方向。
根据血管中心线的坐标拟合出三维B样条曲线后，并求得经过血管中心点的法平面。具体的，以血管中心点为原点，以原数据中y轴方向正交的方向定义为法平面的x轴，与x轴正交的方向为法平面的y轴，以法平面垂直的方向为z轴。
S3.基于法平面对血管壁影像进行采样以获得切片。
在不丢失血管壁所在像素的前提下，选定一个尽量小的采样范围。如在0.33mm/px的成像条件下，可以选择以原点为中心64x64大小的采样范围。在每一个血管中心点的法平面上根据采样范围（可变）获得一个切片，称为一个切片。
S4.基于血管中心线各点的顺序将各切片沿z轴叠加，得到曲面重建的结果。
本申请实施例一种基于中心线提取的血管壁影像自动曲面重建方法的实施原理为：
由深度学习模型输出的血管中心线是由一系列（三维的）点组成的，对其进行拟合为直线以用于表征血管的走向。由于血管得走势迂回，在单个切面上难以显示其全貌，甚至会产生多个端面，经过曲面重建可以将血管拉直，可以反映血管整体的特征。因此，利用血管中心点和血管中心线获取法平面上的切片，再将切片沿z轴叠加，即完成曲面重建和将血管拉直的目的。
本申请实施例还公开一种电子设备，包括存储器和处理器，所述存储器上存储有能够被处理器加载并执行如上述用于黑血MR影像血管壁三维重建的血管中心线提取方法、和/或基于中心线提取的血管壁影像自动曲面重建方法的计算机程序。本实施例方法的执行主体可以是一种控制装置，该控制装置设置在电子设备上，当前设备可以是具有WIFI功能的手机，平板电脑，笔记本电脑等电子设备，本实施例方法的执行主体也可以直接是电子设备的CPU(central processing unit，中央处理器)。
本申请实施例还公开一种计算机可读存储介质，存储有能够被处理器加载并执行如上的用于黑血MR影像血管壁三维重建的血管中心线提取方法、和/或基于中心线提取的血管壁影像自动曲面重建方法的计算机程序。通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件，但很多情况下前者是更佳的实施方式。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如上的一个存储介质(如ROM/RAM、磁碟、光盘)中，包括若干指令用以使得一台设备(可以是手机，计算机，服务器，被控终端，或者网络设备等)执行本申请每个实施例的方法。
以上均为本申请的较佳实施例，并非依此限制本申请的保护范围，故：凡依本申请的结构、形状、原理所做的等效变化，均应涵盖于本申请的保护范围之内。

Claims (10)

  1. 一种用于黑血MR影像血管壁三维重建的血管中心线提取方法，其特征在于，包括以下步骤：
    获取待检测影像；其中，所述的待检测影像为黑血MR影像；
    基于第一目标检测模型从将待检测影像中获取第一血管特征，其中，所述第一目标检测模型被训练用于从二维影像中提取特征，第一血管特征用于粗略对应于血管中心线在待检测影像上的位置；
    叠加第一血管特征至待检测影像并形成两个通道数据；
    将叠加有第一血管特征的待检测影像送入三维影像分割模型，以获得第二血管特征，其中，所述三维影像分割模型用于基于三维卷积核提取三维特征信息，第二血管特征用于精细对应于血管中心线在待检测影像上的三维空间位置；
    基于第二目标检测模型和第二血管特征获得血管中心线坐标。
  2. 根据权利要求1所述的血管中心线提取方法，其特征在于，所述第一血管特征为六个维度确定的N个矩形框，其中，N为待检测影像中出现的第N个目标，X为预测出的目标中心的x轴坐标，Y为预测出的目标中心的y轴坐标，W为预测出的目标的宽度，H为预测出的目标的高度，C为识别到的目标的类别。
  3. 根据权利要求1所述的血管中心线提取方法，其特征在于，所述第一目标检测模型和第二目标检测模型为Yolo模型、SSD模型、Mask-R-CNN模型或Fast-R-CNN模型；
    和/或，所述三维影像分割模型为V-net模型。
  4. 一种基于中心线提取的血管壁影像自动曲面重建方法，其特征在于，包括以下步骤：
    基于如权利要求1-3任意一项所述的血管中心线提取方法获取血管中心点的坐标并拟合；
    基于拟合结果计算经过各血管中心点的法平面，并基于各法平面分别建立空间坐标系，其中z轴为与法平面的法线方向；
    基于法平面对血管壁影像进行采样以获得切片；
    基于血管中心线各点的顺序将各切片沿z轴叠加，得到曲面重建的结果。
  5. 根据权利要求4所述的血管壁影像自动曲面重建方法，其特征在于，所述的拟合血管中心线的各点坐标的方法为B样条拟合。
  6. 根据权利要求4所述的血管壁影像自动曲面重建方法，其特征在于，在基于法平面对血管壁影像进行采样以获得切片之前，还包括以下步骤：
    基于血管类型选取采样范围。
  7. 一种用于黑血MR影像血管壁三维重建的血管中心线提取系统，其特征在于，包括：
    输入模块，用于获取待检测影像；其中，所述的待检测影像为黑血MR影像；
    第一检测模块，用于基于Yolo模型从将待检测影像中获取第一血管特征，其中，第一血管特征用于粗略对应于血管中心线在待检测影像上的位置；
    组合模块，用于叠加第一血管特征至待检测影像并形成两个通道数据；
    第二检测模块，用于将叠加有第一血管特征的待检测影像送入V-net模型，以获得第二血管特征，其中，第二血管特征用于精细对应于血管中心线在待检测影像上的位置；
    第三检测模块，用于基于Yolo网络和第二血管特征获得血管中心线坐标。
  8. 一种基于中心线提取的血管壁影像自动曲面重建系统，其特征在于，包括：
    如权利要求7所述的血管中心线提取系统，用于获取血管中心线的各点坐标；
    拟合模块，用于拟合血管中心线的各点坐标；
    建模模块，用于基于拟合结果计算经过各血管中心点的法平面，并基于各法平面分别建立空间坐标系，其中z轴为与法平面的法线方向；
    采样模块，用于基于法平面对血管壁影像进行采样以获得切片；
    重建模块，用于基于血管中心线各点的顺序将各切片沿z轴叠加，得到曲面重建的结果。
  9. 一种电子设备，其特征在于，包括：
    一个或多个处理器；
    存储器；
    一个或多个应用程序，其中所述一个或多个应用程序被存储在所述存储器中并被配置为由所述一个或多个处理器执行，所述一个或多个程序配置用于：
    执行根据权利要求1至3任一项所述的血管中心线提取方法，
    和/或，执行根据权利要求4至6任一项所述的血管壁影像自动曲面重建方法。
  10. 一种计算机可读存储介质，其特征在于，所述存储介质存储有至少一条指令、至少一段程序、代码集或指令集，所述至少一条指令、所述至少一段程序、所述代码集或指令集由所述处理器加载并执行以实现：
    如权利要求1至3任一项所述的血管中心线提取方法；
    如权利要求4至6任一项所述的血管壁影像自动曲面重建方法。
PCT/CN2022/136953 2021-12-28 2022-12-06 一种基于中心线提取的血管壁影像自动曲面重建方法 WO2023124830A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111621328.1A CN114399594A (zh) 2021-12-28 2021-12-28 一种基于中心线提取的血管壁影像自动曲面重建方法
CN202111621328.1 2021-12-28

Publications (1)

Publication Number Publication Date
WO2023124830A1 true WO2023124830A1 (zh) 2023-07-06

Family

ID=81229423

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/136953 WO2023124830A1 (zh) 2021-12-28 2022-12-06 一种基于中心线提取的血管壁影像自动曲面重建方法

Country Status (2)

Country Link
CN (1) CN114399594A (zh)
WO (1) WO2023124830A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116758050A (zh) * 2023-07-12 2023-09-15 强联智创（北京）科技有限公司 用于颅内Wills环血管中心线盲补全的方法及产品
CN118334273A (zh) * 2024-06-12 2024-07-12 柏意慧心(杭州)网络科技有限公司 一种主动脉窦管交界平面自动提取方法、装置及存储介质

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114399594A (zh) * 2021-12-28 2022-04-26 深圳先进技术研究院 一种基于中心线提取的血管壁影像自动曲面重建方法
CN115456950B (zh) * 2022-08-15 2023-08-11 上海交通大学医学院附属第九人民医院 一种穿支血管自动识别和定位方法、设备及存储介质
CN116934741B (zh) * 2023-09-11 2023-12-26 首都医科大学附属北京天坛医院 一站式血管壁的成分和定量参数获取方法和装置

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110150274A1 (en) * 2009-12-23 2011-06-23 General Electric Company Methods for automatic segmentation and temporal tracking
CN111681226A (zh) * 2020-06-09 2020-09-18 上海联影医疗科技有限公司 基于血管识别的目标组织定位方法和装置
CN112561781A (zh) * 2020-12-07 2021-03-26 深圳先进技术研究院 一种磁共振血管壁图像分析方法、系统及计算机可读介质
CN113223015A (zh) * 2021-05-11 2021-08-06 清华大学 血管壁图像分割方法、装置、计算机设备和存储介质
CN114399594A (zh) * 2021-12-28 2022-04-26 深圳先进技术研究院 一种基于中心线提取的血管壁影像自动曲面重建方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110150274A1 (en) * 2009-12-23 2011-06-23 General Electric Company Methods for automatic segmentation and temporal tracking
CN111681226A (zh) * 2020-06-09 2020-09-18 上海联影医疗科技有限公司 基于血管识别的目标组织定位方法和装置
CN112561781A (zh) * 2020-12-07 2021-03-26 深圳先进技术研究院 一种磁共振血管壁图像分析方法、系统及计算机可读介质
CN113223015A (zh) * 2021-05-11 2021-08-06 清华大学 血管壁图像分割方法、装置、计算机设备和存储介质
CN114399594A (zh) * 2021-12-28 2022-04-26 深圳先进技术研究院 一种基于中心线提取的血管壁影像自动曲面重建方法

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116758050A (zh) * 2023-07-12 2023-09-15 强联智创（北京）科技有限公司 用于颅内Wills环血管中心线盲补全的方法及产品
CN118334273A (zh) * 2024-06-12 2024-07-12 柏意慧心(杭州)网络科技有限公司 一种主动脉窦管交界平面自动提取方法、装置及存储介质

Also Published As

Publication number Publication date
CN114399594A (zh) 2022-04-26

Similar Documents

Publication Publication Date Title
US12067725B2 (en) Image region localization method, image region localization apparatus, and medical image processing device
WO2023124830A1 (zh) 一种基于中心线提取的血管壁影像自动曲面重建方法
Yang et al. Robust segmentation of arterial walls in intravascular ultrasound images using Dual Path U-Net
Liu et al. Deep convolutional neural network and 3D deformable approach for tissue segmentation in musculoskeletal magnetic resonance imaging
CN107997778A (zh) 在计算机断层扫描血管造影术中基于深度学习的骨移除
Kang et al. ASF-YOLO: A novel YOLO model with attentional scale sequence fusion for cell instance segmentation
Shi et al. Automatic segmentation of cardiac magnetic resonance images based on multi-input fusion network
CN110036408A (zh) 活动性出血和血液外渗的自动ct检测和可视化
Venkatesh et al. Intensity inhomogeneity correction of MRI images using InhomoNet
Chai et al. MRI restoration using edge-guided adversarial learning
CN109754388B (zh) 一种颈动脉狭窄程度计算方法、装置及存储介质
Zou et al. Multi-scale deformable transformer for multi-contrast knee MRI super-resolution
CN107516314B (zh) 医学图像超体素分割方法和装置
Hameeteman et al. Carotid wall volume quantification from magnetic resonance images using deformable model fitting and learning-based correction of systematic errors
CN110136107B (zh) 基于dssd和时域约束x光冠脉造影序列自动分析方法
Wang et al. Automatic consecutive context perceived transformer GAN for serial sectioning image blind inpainting
Tong et al. Automatic lumen border detection in IVUS images using dictionary learning and kernel sparse representation
Aja-Fernández et al. Validation of deep learning techniques for quality augmentation in diffusion MRI for clinical studies
Lin et al. Method for carotid artery 3-D ultrasound image segmentation based on cswin transformer
Fourcade et al. Deformable image registration with deep network priors: a study on longitudinal PET images
US8170321B2 (en) System and method for contour tracking in cardiac phase contrast flow MR images
Zhang et al. 3d cross-scale feature transformer network for brain mr image super-resolution
Yu et al. Segmentation of cardiac tagged MR images using a snake model based on hybrid gradient vector flow
CN113450394B (zh) 一种基于Siamese网络的异尺寸图像配准方法
Zeineldin et al. Self-supervised iRegNet for the Registration of Longitudinal Brain MRI of Diffuse Glioma Patients

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: 22914085

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE