CN111612778A - Preoperative CTA and intraoperative X-ray coronary artery registration method - Google Patents

Preoperative CTA and intraoperative X-ray coronary artery registration method Download PDF

Info

Publication number
CN111612778A
CN111612778A CN202010456226.8A CN202010456226A CN111612778A CN 111612778 A CN111612778 A CN 111612778A CN 202010456226 A CN202010456226 A CN 202010456226A CN 111612778 A CN111612778 A CN 111612778A
Authority
CN
China
Prior art keywords
ray
cta
blood vessel
points
intraoperative
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202010456226.8A
Other languages
Chinese (zh)
Other versions
CN111612778B (en
Inventor
顾力栩
张宏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai Jiaotong University
Original Assignee
Shanghai Jiaotong University
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 Shanghai Jiaotong University filed Critical Shanghai Jiaotong University
Priority to CN202010456226.8A priority Critical patent/CN111612778B/en
Publication of CN111612778A publication Critical patent/CN111612778A/en
Application granted granted Critical
Publication of CN111612778B publication Critical patent/CN111612778B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
    • G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
    • A61B6/02—Arrangements for diagnosis sequentially in different planes; Stereoscopic radiation diagnosis
    • A61B6/03—Computed tomography [CT]
    • A61B6/032—Transmission computed tomography [CT]
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
    • A61B6/50—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications
    • A61B6/504—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications for diagnosis of blood vessels, e.g. by angiography
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
    • A61B6/52—Devices using data or image processing specially adapted for radiation diagnosis
    • A61B6/5211—Devices using data or image processing specially adapted for radiation diagnosis involving processing of medical diagnostic data
    • A61B6/5217—Devices using data or image processing specially adapted for radiation diagnosis involving processing of medical diagnostic data extracting a diagnostic or physiological parameter from medical diagnostic data
    • 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
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/66—Analysis of geometric attributes of image moments or centre of gravity
    • 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
    • G06T2207/10116—X-ray image
    • 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
    • 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
    • G06T2207/30172—Centreline of tubular or elongated structure

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Medical Informatics (AREA)
  • Theoretical Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Biophysics (AREA)
  • Molecular Biology (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Animal Behavior & Ethology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Pathology (AREA)
  • Surgery (AREA)
  • Optics & Photonics (AREA)
  • Public Health (AREA)
  • High Energy & Nuclear Physics (AREA)
  • Veterinary Medicine (AREA)
  • Evolutionary Computation (AREA)
  • Physiology (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Artificial Intelligence (AREA)
  • Pulmonology (AREA)
  • Vascular Medicine (AREA)
  • Dentistry (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Geometry (AREA)
  • Quality & Reliability (AREA)
  • Apparatus For Radiation Diagnosis (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a preoperative CTA (computed tomography angiography) and intraoperative X-ray coronary artery registration method, which relates to the technical field of medical image processing and is characterized in that real-time registration is realized based on a point cloud registration network. The method directly predicts the deformation field through the trained registration network without iterative optimization, thereby meeting the requirement of real-time property, having the advantages of rapidness, accuracy, suitability for practical application and the like, and reducing the complexity of the operation of a doctor.

Description

Preoperative CTA and intraoperative X-ray coronary artery registration method
Technical Field
The invention relates to the technical field of medical image processing, in particular to a preoperative CTA and intraoperative X-ray coronary artery registration method.
Background
In recent years, with the improvement of living standard of people, the incidence rate of cardiovascular diseases is higher and higher. The currently common treatment is an interventional procedure, which aims to open a stenosis with a guide wire. During interventional procedures, physicians often determine the position and direction of travel of a guidewire by taking radiographic images. However, since X-ray coronary angiography has different degrees of artifacts, it presents a great challenge to the surgeon's procedure. Meanwhile, the success rate of the operation depends on the actual experience of the doctor to a great extent. Before surgery, physicians have a rough understanding of a patient's blood vessels by taking CTAs, but often consider them separately from intraoperative X-rays.
The search of the prior art finds that the coronary artery registration method commonly used at present is a feature-based registration method. Such as the most recent iterative point method, which treats the coronary registration problem as a rigid point set registration problem. Firstly, extracting a vessel central line as a point set to be registered, and then iteratively searching an optimal rigid transformation matrix by optimizing the Euclidean distance between the two point sets. The consistency point drift method extends the coronary registration to the non-rigid registration problem, which is considered to be a probability density estimation problem. It fits the center of one set of points to another set of points by likelihood maximization. However, since the above methods all require iterative optimization, it is often difficult to meet the real-time requirement. With the development of deep learning in the field of image processing, point cloud registration methods based on learning are also greatly developed. The method utilizes a point cloud network to extract features from a point set, inputs the features into a multilayer perceptron and outputs required transformation parameters. Such methods can predict the deformation field end-to-end, which also provides a solution for coronary registration.
Accordingly, those skilled in the art have sought to develop methods for real-time registration of preoperative CTA with intraoperative X-rays to reduce the complexity of the physician procedure.
Disclosure of Invention
In view of the above problems, the present invention aims to realize real-time registration of preoperative CTA and intraoperative X-ray, and adopts an end-to-end preoperative CTA and intraoperative X-ray coronary artery registration method. Through the trained registration network, the deformation field is directly predicted without iterative optimization, so that the requirement of real-time performance is met.
In order to achieve the aim, the invention provides a preoperative CTA and intraoperative X-ray coronary artery registration method which is characterized in that real-time registration is realized based on a point cloud registration network.
Further, the preoperative CTA and intraoperative X-ray coronary artery registration method comprises the following steps:
step 1, performing blood vessel segmentation on preoperative CTA data and intraoperative X-ray data by using a convolutional neural network U-Net to obtain a CTA blood vessel and an X-ray blood vessel, extracting a maximum communication area, and performing center line extraction on a segmentation result by using a skeleton extraction algorithm to obtain a CTA center line and an X-ray center line;
step 2, calibrating the X-ray machine, reading DICOM header file information, constructing a projection matrix, and generating a projection center line by using the CTA center line;
and 3, performing point set sampling on the projection central line and the X-ray central line, inputting the point set sampling into the point cloud registration network to obtain a deformation field, deforming the projection central line by using the deformation field, calculating a loss function with the X-ray central line, and optimizing the point cloud registration network.
Further, the step 1 comprises the following steps:
step 1.1, performing pixel-by-pixel blood vessel labeling on the preoperative CTA data, and training the convolution neural network U-Net for segmenting the CTA blood vessel by using an original image and a corresponding label;
step 1.2, performing pixel-by-pixel blood vessel labeling on the intraoperative X-ray data, and training the convolution neural network U-Net for the X-ray blood vessel segmentation by using the original image and a corresponding label;
step 1.3, performing the maximum connected domain analysis on the CTA blood vessel and the X-ray blood vessel obtained by the segmentation, reserving the maximum connected domain, extracting a main blood vessel and removing noise;
and 1.4, extracting the central line of the blood vessel by using the skeleton extraction algorithm.
Further, step 2 reads the DICOM header file information, where the DICOM header file information includes: (1) the probe to target distance SDD; (2) the width W of the image; (3) the height H of the image.
Further, the formula for constructing the projection matrix in step 2 is as follows:
Figure BDA0002509333980000021
wherein s isHIs the pixel spacing, s, in the X-axis directionWIs the pixel spacing in the Y-axis direction.
Further, the projection centerline is generated in step 2 using the projection matrix and the CTA centerline, and the formula is as follows:
Figure BDA0002509333980000022
wherein (x)i,yi) Coordinates for each pixel point of the CTA centerline.
Further, the step 3 comprises the following steps:
step 3.1, constructing the point cloud registration network on the basis of the DGCNN point cloud network, wherein the point cloud registration network comprises 5 side convolution layers, the number of convolution kernels of each side convolution layer is 64, 64, 128, 256 and 512, and then 1 three layers of sensing machines are connected, and the number of units of each layer of sensing machine is 256, 128 and 2;
and 3.2, forming a registration pair according to the projection central line generated in the step 2 and the X-ray central line obtained in the step 1. Sampling 600 points of the projection central line and the X-ray central line at random as vessel characteristic points and inputting the vessel characteristic points into the point cloud registration network;
step 3.3, outputting the deformation field through processing the point cloud registration network, wherein the deformation field comprises the displacement of each blood vessel characteristic point and is represented by T;
and 3.4, deforming the projection center line according to the deformation field, wherein the formula is as follows:
Pw(xi,yi)＝Ps(xi,yi)+T(xi,yi)
wherein, Ps(xi,yi) Denotes the coordinates of the ith point, T (x)i,yi) Indicates the deformation of the ith point, Pw(xi,yi) Representing the coordinates of the ith point after deformation;
step 3.5, calculating the loss function of the deformed projection center line and the X-ray center line, wherein the loss function comprises three parts, namely a chamfer angle distance LchamferGlobal topological constraint term LGMMLocal topological constraint term LLPCThe formula is as follows:
L＝LchamferαLGMM+βLLPC
wherein, alpha and beta are used for measuring the strength of the constraint;
and 3.6, reversely propagating and optimizing the point cloud registration network by the calculated loss function value so as to fit the point cloud registration network to an optimal solution.
Further, the chamfer distance LchamferFor measuring the distance between two sets of points that do not correspond, for each point in the first set of points, the closest point in the other set of points is calculated, and then the squares of the distances are summed, as follows:
Figure BDA0002509333980000031
wherein, PwRepresenting the deformed set of points, PtRepresenting a set of target points.
Further, the global topological constraint term LGMMDefined as the gaussian mixture model error between two sets of points, which measures the similarity between two sets of points as a continuous density function, moving one set of points coherently to the other, the formula is as follows:
Figure BDA0002509333980000032
wherein, PwRepresenting the deformed set of points, PtRepresenting a set of target points. The standard deviation of the gaussian mixture model is indicated.
Further, the local topological constraint term LLPCDefined as the gradient between two points of the blood vessel, controlling the length of the blood vesselDegree transformation range, the formula is as follows:
Figure BDA0002509333980000033
wherein, PsRepresenting a set of source points, PwRepresenting the deformed set of points.
Compared with the prior art, the method has the advantages that the deformation field is predicted end to end by utilizing the point cloud registration network, and topological constraint is applied, so that the deformation field is smoother and more reasonable, real-time coronary artery registration is realized, and the method is rapid, accurate, suitable for practical application and the like.
The conception, the specific structure and the technical effects of the present invention will be further described with reference to the accompanying drawings to fully understand the objects, the features and the effects of the present invention.
Drawings
FIG. 1 is a flow chart of the present invention;
fig. 2 is a structural diagram of a point cloud registration network.
Detailed Description
The technical contents of the preferred embodiments of the present invention will be more clearly and easily understood by referring to the drawings attached to the specification. The present invention may be embodied in many different forms of embodiments and the scope of the invention is not limited to the embodiments set forth herein.
The embodiment is realized on a 64-bit Linux system, a 64GB RAM, an Intel (R) Xeon E5-2630 v3 CPU and an Nvidia Titan X passacal GPU computer, and the whole method is realized on the basis of a PyTorch deep learning framework and by adopting a python programming language.
The implementation flow of this embodiment is shown in fig. 1, and specifically includes the following steps:
step 1, performing blood vessel segmentation on preoperative CTA (computed tomography angiography) and intraoperative X-ray data by using a U-Net network, extracting a maximum communication area, and performing center line extraction on a segmentation result by using a skeleton extraction algorithm;
step 2, calibrating the X-ray machine, reading DICOM header file information, constructing a projection matrix, and generating a projection center line by using a CTA center line;
step 3, sampling point sets of the projection center line and the X-ray center line, inputting the point sets into a point cloud registration network to obtain a deformation field, deforming the projection center line by using the deformation field, calculating a loss function with the X-ray center line, and optimizing the point cloud registration network;
further, the step 1 comprises the following steps:
step 1.1, performing pixel-by-pixel blood vessel labeling on preoperative CTA data, and training a convolution neural network U-Net for CTA blood vessel segmentation by using an original image and a corresponding label;
step 1.2, performing pixel-by-pixel blood vessel labeling on the X-ray data in the operation, and training a convolution neural network U-Net for X-ray blood vessel segmentation by using an original image and a corresponding label;
step 1.3, performing maximum connected domain analysis on the CTA and the X-ray blood vessel obtained by the segmentation, reserving the maximum connected domain, extracting a main blood vessel, and removing noise;
and 1.4, extracting the center line of the blood vessel by using a skeleton extraction algorithm.
Further, the step 2 comprises the following steps:
step 2.1, calibrating the X-ray machine, and reading DICOM header file information, wherein the information comprises the following contents: (1) the probe to target distance SDD; (2) the width W of the image; (3) the height H of the image. The projection matrix is constructed with the following definitions:
Figure BDA0002509333980000051
wherein s isHIs the pixel spacing, s, in the X-axis directionWIs the pixel spacing in the Y-axis direction.
Step 2.2 generates a projection centerline using the projection matrix and the CTA centerline. The specific formula is as follows:
Figure BDA0002509333980000052
wherein (x)i,yi) Coordinates for each pixel point of the CTA centerline.
Further, the step 3 comprises the following steps:
and 3.1, building a registration network on the basis of the DGCNN point cloud network, wherein the registration network comprises 5 side convolution layers, and the number of convolution kernels of each side convolution layer is 64, 64, 128, 256 and 512 respectively. And then, connecting a three-layer perceptron, wherein the unit number of each perceptron is respectively 256, 128 and 2.
And 3.2, forming a registration pair according to the projection central line generated in the step 2 and the X-ray central line obtained in the step 1. Two centerlines are sampled at random for 600 points and input into the network as vessel feature points.
And 3.3, outputting a deformation field through the processing of the registration network. The deformation field includes the displacement of each feature point, denoted by T.
And 3.4, deforming the projection center line according to the deformation field, wherein the specific formula is as follows:
Pw(xi,yi)＝Ps(xi,yi)+T(xi,yi)
wherein, Ps(xi,yi) Denotes the coordinates of the ith point, T (x)i,yi) Indicates the deformation of the ith point, Pw(xi,yi) And (4) representing the coordinates of the ith point after deformation.
And 3.5, calculating a loss function of the deformed central line and the X-ray central line. The loss function includes three parts, namely a chamfer distance LchamferGlobal topological constraint term LGMMLocal topological constraint term LLPC. The specific formula is as follows:
L＝LchamferαLGMM+βLLPC
where α and β are used to measure the strength of the constraint.
Chamfer distance LchamferTo measure the distance between two sets of points that do not correspond. For each point in the first set of points, the closest point in the other set of points is calculated and the squares of the distances are then summed up, as defined below:
Figure BDA0002509333980000053
wherein, PwRepresenting the deformed set of points, PtRepresenting a set of target points.
Global topological constraint term LGMMThe method is defined as a Gaussian mixture model error between two point sets, measures similarity between the two point sets as a continuous density function, and moves one point set to the other point set in a coherent mode, and is specifically defined as follows:
Figure BDA0002509333980000062
wherein, PwRepresenting the deformed set of points, PtRepresenting a set of target points. The standard deviation of the gaussian mixture model is indicated.
Local topological constraint term LLPCThe gradient between two points of the blood vessel is defined, so that the length transformation range of the blood vessel is not too large, and the gradient is specifically defined as follows:
Figure BDA0002509333980000061
wherein, PsRepresenting a set of source points, PwRepresenting the deformed set of points.
And 3.6, reversely propagating and optimizing the point cloud registration network by the calculated loss function value so as to fit the point cloud registration network to an optimal solution.
The foregoing detailed description of the preferred embodiments of the invention has been presented. It should be understood that numerous modifications and variations could be devised by those skilled in the art in light of the present teachings without departing from the inventive concepts. Therefore, the technical solutions available to those skilled in the art through logic analysis, reasoning and limited experiments based on the prior art according to the concept of the present invention should be within the scope of protection defined by the claims.

Claims (10)

1. A preoperative CTA and intraoperative X-ray coronary artery registration method is characterized in that real-time registration is achieved based on a point cloud registration network.
2. The preoperative CTA and intraoperative X-ray coronary registration method of claim 1, comprising the steps of:
step 1, performing blood vessel segmentation on preoperative CTA data and intraoperative X-ray data by using a convolutional neural network U-Net to obtain a CTA blood vessel and an X-ray blood vessel, extracting a maximum communication area, and performing center line extraction on a segmentation result by using a skeleton extraction algorithm to obtain a CTA center line and an X-ray center line;
step 2, calibrating the X-ray machine, reading DICOM header file information, constructing a projection matrix, and generating a projection center line by using the CTA center line;
and 3, performing point set sampling on the projection central line and the X-ray central line, inputting the point set sampling into the point cloud registration network to obtain a deformation field, deforming the projection central line by using the deformation field, calculating a loss function with the X-ray central line, and optimizing the point cloud registration network.
3. The preoperative CTA and intraoperative X-ray coronary registration method of claim 2, wherein said step 1 comprises the steps of:
step 1.1, performing pixel-by-pixel blood vessel labeling on the preoperative CTA data, and training the convolution neural network U-Net for segmenting the CTA blood vessel by using an original image and a corresponding label;
step 1.2, performing pixel-by-pixel blood vessel labeling on the intraoperative X-ray data, and training the convolution neural network U-Net for the X-ray blood vessel segmentation by using the original image and a corresponding label;
step 1.3, performing the maximum connected domain analysis on the CTA blood vessel and the X-ray blood vessel obtained by the segmentation, reserving the maximum connected domain, extracting a main blood vessel and removing noise;
and step 1.4, performing the centerline extraction on the CTA blood vessel and the X-ray blood vessel by using the skeleton extraction algorithm to obtain the CTA centerline and the X-ray centerline.
4. The method of pre-operative CTA and intra-operative X-ray coronary registration of claim 2, wherein said step 2 reads the DICOM header file information, which comprises: (1) the probe to target distance SDD; (2) the width W of the image; (3) the height H of the image.
5. The preoperative CTA and intraoperative X-ray coronary registration method of claim 4, wherein said projection matrix constructed in step 2 is formulated as:
Figure FDA0002509333970000011
wherein s isHIs the pixel spacing, s, in the X-axis directionWIs the pixel spacing in the Y-axis direction.
6. The preoperative CTA and intraoperative X-ray coronary registration method of claim 2, wherein said projection centerline is generated in said step 2 using said projection matrix and said CTA centerline as follows:
Figure FDA0002509333970000021
wherein (x)i，yi) Coordinates for each pixel point of the CTA centerline.
7. The preoperative CTA and intraoperative X-ray coronary registration method of claim 2, wherein said step 3 comprises the steps of:
step 3.1, constructing the point cloud registration network on the basis of the DGCNN point cloud network, wherein the point cloud registration network comprises 5 side convolution layers, the number of convolution kernels of each side convolution layer is 64, 64, 128, 256 and 512 respectively, and 1 three layers of sensing machines are connected, and the unit number of each layer of sensing machine is 256, 128 and 2 respectively;
step 3.2, forming a registration pair according to the projection central line generated in the step 2 and the X-ray central line obtained in the step 1, and randomly sampling 600 points of the projection central line and the X-ray central line to be used as blood vessel characteristic points to be input into the point cloud registration network;
step 3.3, outputting the deformation field through processing the point cloud registration network, wherein the deformation field comprises the displacement of each blood vessel characteristic point and is represented by T;
and 3.4, deforming the projection center line according to the deformation field, wherein the formula is as follows:
Pw(xi，yi)＝Ps(xi，yi)+T(xi，yi)
wherein, Ps(xi，yi) Denotes the coordinates of the ith point, T (x)i，yi) Indicates the deformation of the ith point, Pw(xi，yi) Representing the coordinates of the ith point after deformation;
step 3.5, calculating the loss function of the deformed projection center line and the X-ray center line, wherein the loss function comprises three parts, namely a chamfer angle distance LchamferGlobal topological constraint term LGMMLocal topological constraint term LLPCThe formula is as follows:
L＝Lchamfer+αLGMM+βLLPC
wherein, alpha and beta are used for measuring the strength of the constraint;
and 3.6, reversely propagating and optimizing the point cloud registration network by the calculated loss function value so as to fit the point cloud registration network to an optimal solution.
8. The preoperative CTA and intraoperative X-ray coronary registration method of claim 7,
the chamfer distance LchamferFor measuring the distance between two sets of points that do not correspond, for each point in the first set of points, the closest point in the other set of points is calculated, and then the squares of the distances are summed, as follows:
Figure FDA0002509333970000022
wherein, PwRepresenting the deformed set of points, PtRepresenting a set of target points.
9. The preoperative CTA and intraoperative X-ray coronary registration method of claim 7, wherein said global topological constraint term LGMMDefined as the gaussian mixture model error between two sets of points, which measures the similarity between two sets of points as a continuous density function, moving one set of points coherently to the other, the formula is as follows:
Figure FDA0002509333970000031
wherein, PwRepresenting the deformed set of points, PtRepresenting a set of target points. The standard deviation of the gaussian mixture model is indicated.
10. The preoperative CTA and intraoperative X-ray coronary registration method of claim 7, wherein said local topological constraint term LLPCThe gradient between two points of the blood vessel is defined, and the length transformation range of the blood vessel is controlled, and the formula is as follows:
Figure FDA0002509333970000032
wherein, PsRepresenting a set of source points, PwRepresenting the deformed set of points.
CN202010456226.8A 2020-05-26 2020-05-26 Preoperative CTA and intraoperative X-ray coronary artery registration method Active CN111612778B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010456226.8A CN111612778B (en) 2020-05-26 2020-05-26 Preoperative CTA and intraoperative X-ray coronary artery registration method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010456226.8A CN111612778B (en) 2020-05-26 2020-05-26 Preoperative CTA and intraoperative X-ray coronary artery registration method

Publications (2)

Publication Number Publication Date
CN111612778A true CN111612778A (en) 2020-09-01
CN111612778B CN111612778B (en) 2023-07-11

Family

ID=72196340

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010456226.8A Active CN111612778B (en) 2020-05-26 2020-05-26 Preoperative CTA and intraoperative X-ray coronary artery registration method

Country Status (1)

Country Link
CN (1) CN111612778B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112991407A (en) * 2021-04-02 2021-06-18 浙江大学计算机创新技术研究院 Point cloud registration method based on non-local operation
CN114972165A (en) * 2022-03-24 2022-08-30 中山大学孙逸仙纪念医院 Method and device for measuring time-average shearing force
WO2022194258A1 (en) * 2021-03-17 2022-09-22 先临三维科技股份有限公司 Method and apparatus for training dental cast deformation model
TWI797767B (en) * 2021-10-06 2023-04-01 中原大學 Device and method of angiography for cerebrovascular obliteration

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102999902A (en) * 2012-11-13 2013-03-27 上海交通大学医学院附属瑞金医院 Optical navigation positioning system based on CT (computed tomography) registration results and navigation method thereby
CN106097334A (en) * 2016-06-08 2016-11-09 北京航空航天大学 Point cloud registration method based on degree of depth study
CN109300147A (en) * 2018-09-17 2019-02-01 北京理工大学 The method and device of blood vessel two-dimensional center line and three-dimensional center line registration
CN109741344A (en) * 2018-12-28 2019-05-10 上海联影智能医疗科技有限公司 Blood vessel segmentation method, apparatus, medical imaging equipment and storage medium
CN110009669A (en) * 2019-03-22 2019-07-12 电子科技大学 A kind of 3D/2D medical image registration method based on deeply study

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102999902A (en) * 2012-11-13 2013-03-27 上海交通大学医学院附属瑞金医院 Optical navigation positioning system based on CT (computed tomography) registration results and navigation method thereby
CN106097334A (en) * 2016-06-08 2016-11-09 北京航空航天大学 Point cloud registration method based on degree of depth study
CN109300147A (en) * 2018-09-17 2019-02-01 北京理工大学 The method and device of blood vessel two-dimensional center line and three-dimensional center line registration
CN109741344A (en) * 2018-12-28 2019-05-10 上海联影智能医疗科技有限公司 Blood vessel segmentation method, apparatus, medical imaging equipment and storage medium
CN110009669A (en) * 2019-03-22 2019-07-12 电子科技大学 A kind of 3D/2D medical image registration method based on deeply study

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
GAURAV KAILA等: "Fusion of CTA and XA data using 3D centerline registration for plaque visualization during coronary intervention" *
HAOFU LIAO等: "Multiview 2D/3D Rigid Registration via a Point-Of-Interest Network for Tracking and Triangulation" *
李致勋: "2D/3D 冠状动脉血管分割与配准方法研究" *
李致勋: "2D/3D冠状动脉血管分割与配准方法研究" *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2022194258A1 (en) * 2021-03-17 2022-09-22 先临三维科技股份有限公司 Method and apparatus for training dental cast deformation model
CN112991407A (en) * 2021-04-02 2021-06-18 浙江大学计算机创新技术研究院 Point cloud registration method based on non-local operation
TWI797767B (en) * 2021-10-06 2023-04-01 中原大學 Device and method of angiography for cerebrovascular obliteration
CN114972165A (en) * 2022-03-24 2022-08-30 中山大学孙逸仙纪念医院 Method and device for measuring time-average shearing force
CN114972165B (en) * 2022-03-24 2024-03-15 中山大学孙逸仙纪念医院 Method and device for measuring time average shearing force

Also Published As

Publication number Publication date
CN111612778B (en) 2023-07-11

Similar Documents

Publication Publication Date Title
US11657330B2 (en) System for guiding medically invasive devices relative to anatomical structures via image processing
CN111612778B (en) Preoperative CTA and intraoperative X-ray coronary artery registration method
JP7391846B2 (en) Computer-aided diagnosis using deep neural networks
US20210192758A1 (en) Image processing method and apparatus, electronic device, and computer readable storage medium
Ambrosini et al. Fully automatic and real-time catheter segmentation in X-ray fluoroscopy
JP6243453B2 (en) Multimodal segmentation in intravascular images
CN109938836B (en) Automatic unmarked guide wire guiding method in vascular intervention robot operation
CN110110808B (en) Method and device for performing target labeling on image and computer recording medium
Zhang et al. 3D ultrasound centerline tracking of abdominal vessels for endovascular navigation
CN114792326A (en) Surgical navigation point cloud segmentation and registration method based on structured light
Tahoces et al. Deep learning method for aortic root detection
CN110136107B (en) Automatic analysis method based on DSSD and time domain constraint X-ray coronary angiography sequence
Li et al. EchoEFNet: multi-task deep learning network for automatic calculation of left ventricular ejection fraction in 2D echocardiography
CN116612166A (en) Registration fusion algorithm for multi-mode images
Ma et al. PCA-derived respiratory motion surrogates from X-ray angiograms for percutaneous coronary interventions
CN116109872A (en) Blood vessel naming method and device, electronic equipment and storage medium
CN113935889A (en) Method, system and medium for automatic 2D/3D coronary artery registration
Marie et al. Segmentation of kidneys deformed by nephroblastoma using case-based reasoning
Younisse et al. Fine-tuning U-net for medical image segmentation based on activation function, optimizer and pooling layer.
Xia et al. Context region discovery for automatic motion compensation in fluoroscopy
Zhao et al. Development of an approach to extracting coronary arteries and detecting stenosis in invasive coronary angiograms
CN115511773B (en) Blood vessel center line extraction method, device, equipment and readable storage medium
CN115578564B (en) Training method and device for instance segmentation model, electronic equipment and storage medium
CN115482181B (en) Image information extraction method and device, electronic equipment and readable storage medium
US20240153094A1 (en) Systems and methods for annotating tubular structures

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant