CN108010056B - Vascular motion tracking method based on four-dimensional medical image - Google Patents

Vascular motion tracking method based on four-dimensional medical image Download PDF

Info

Publication number
CN108010056B
CN108010056B CN201711205773.3A CN201711205773A CN108010056B CN 108010056 B CN108010056 B CN 108010056B CN 201711205773 A CN201711205773 A CN 201711205773A CN 108010056 B CN108010056 B CN 108010056B
Authority
CN
China
Prior art keywords
blood vessel
volume
time phase
block
dimensional
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.)
Active
Application number
CN201711205773.3A
Other languages
Chinese (zh)
Other versions
CN108010056A (en
Inventor
张辉
齐天卉
杜康
李晓光
卓力
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.)
Beijing University of Technology
Original Assignee
Beijing University of Technology
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 Beijing University of Technology filed Critical Beijing University of Technology
Priority to CN201711205773.3A priority Critical patent/CN108010056B/en
Publication of CN108010056A publication Critical patent/CN108010056A/en
Application granted granted Critical
Publication of CN108010056B publication Critical patent/CN108010056B/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
    • G06T7/20—Analysis of motion
    • G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
    • G06T7/251—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments involving models
    • 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
    • G06T2207/10076—4D tomography; Time-sequential 3D tomography
    • 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
    • G06T2207/30241—Trajectory

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Apparatus For Radiation Diagnosis (AREA)

Abstract

A vessel motion tracking method based on four-dimensional medical images belongs to the field of medical image processing. Firstly, the invention constructs a vascular morphology modeling mode combining layering and chain structure, and realizes the parametric representation of the vascular morphology structure in three-dimensional space; secondly, aiming at the complex motion situation of the blood vessel in the three-dimensional space, a visual description model of the whole blood vessel is established from appearance expression, spatial relation and local structure form of the local blood vessel, and the problem of integral model description of the blood vessel in the volume image is solved; finally, the invention establishes a dynamic deduction method of the state of the blood vessel model, and provides a set of blood vessel positioning integral scheme from three-dimensional image blood vessel description to four-dimensional image blood vessel motion tracking. The invention can perform robust and accurate three-dimensional motion tracking on interested vessels section by section in the whole respiratory or cardiac cycle, the tracking result does not depend on the performance of complex three-dimensional registration operation in the traditional technology, and the whole calculation process is efficient and simple.

Description

Vascular motion tracking method based on four-dimensional medical image
Technical Field
The invention belongs to the technical field of medical image processing, relates to a method for tracking a moving target, and more particularly relates to a four-dimensional medical image-oriented blood vessel motion tracking method.
Background
Blood vessels are the main transportation channels for human body life movement, and have been the focus of attention in the medical field. Clinically, quantitative analysis and measurement are carried out on blood vessel motion caused by respiration and heartbeat motion, and doctors can be assisted in carrying out accurate diagnosis and treatment of diseases. For example, coronary arteries are important pathways for myocardial blood supply, and cardiovascular disease diagnosis and cardiovascular function assessment can be realized by extracting and tracking motion of image data; meanwhile, the operation risk is avoided in the cardiovascular operation and thoracoabdominal radiotherapy stages, and the occurrence of operation complications is reduced. Therefore, it is of great medical value to accurately and robustly locate the position of the blood vessel from the medical image data.
In recent years, with the progress of medical imaging technology, time factors are fused into three-dimensional volume data to form a 3D + T four-dimensional medical image, which is gradually and widely applied in clinical application of precise medicine. Aiming at the cardiovascular imaging requirements of living bodies, the four-dimensional imaging technology can provide dynamic three-dimensional information of a complete heart structure, more accurately reflects the motion range and the space position change rule of each region of the chest and abdomen, obviously reduces respiratory motion artifacts, relatively accurately reflects the displacement and the motion of each blood vessel, can describe the blood vessel motion track of the whole cardiac cycle and respiratory cycle, and improves the blood vessel motion tracking precision and robustness. The four-dimensional imaging technology provides an effective observation means for developing human organ motor energy analysis, but the further analysis of image data and the acquisition of quantitative motion parameters become the problems to be solved urgently in the field of medical image processing.
The vessel motion tracking algorithm is continuously developed under the drive of the innovation of the angiography imaging technology, and the image types can be divided into two-dimensional medical images and three-dimensional medical images. The vessel tracking in the two-dimensional image represents a Snake model method based on single vessel tracking and a landmark method for vessel tree motion. The Snake model method is adopted to minimize energy function and conflict between upper knowledge and bottom image characteristics, can naturally adapt to the target shape and introduce image appearance constraint into blood vessel deformation estimation, and is suitable for tracking the blood vessel structure which is bent in one direction and does not move violently between frames. The T-Snake algorithm in the medical field determines a boundary triangle by means of a triangular mesh and a characteristic function of a mesh point, so that the T-Snake algorithm can process medical images with complex topological structures. The landmark point-based method is used for realizing the structured tracking of the moving vessel tree by taking the vessel bifurcation point as a landmark point for vessel motion tracking. The method calculates the minimum angle difference by taking the 'Y' shape as a mark at the bifurcation of the blood vessel, and mainly solves the problems of volume intensity change and background noise.
For the tracking of the vessel motion under the three-dimensional imaging condition, the prior art focuses on two major ideas, namely a tracking method based on deformation registration and a tracking method based on a vessel structure model. The former scheme focuses on the characteristic of motion local consistency in a three-dimensional space, converts the tracking of sequence data into the registration problem of pairwise time phase data, firstly constructs parameterized blood vessel description by extracting a blood vessel central line, and then solves the deformation parameter of the blood vessel central line under an optimization framework of image registration. The scheme of the latter establishes parameterized or unparameterized vascular structure model description, and realizes the motion parameter estimation of the blood vessels by combining local blood vessel matching or template matching under the constraint of the spatial structure. In general, the typical tracking algorithms of two-dimensional medical images and three-dimensional medical images have the problems of low blood vessel extraction precision, high operation complexity, low tracking robustness and the like.
In view of the above-mentioned technical development, the motion tracking of blood vessels is not yet effectively solved. The invention aims at the problem of tracking the movement of the blood vessel, excavates the effective characteristics of the description blood vessel, constructs a blood vessel description model and establishes a blood vessel movement track tracking technology.
Disclosure of Invention
The invention provides a robust blood vessel motion tracking method aiming at the blood vessel tracking and positioning requirements in medical clinical four-dimensional images and aiming at the conditions of low blood vessel imaging signal-to-noise ratio and complex blood vessel form change in dynamic volume data. The technical means adopted by the invention constructs a blood vessel description model from three aspects of volume intensity information, blood vessel structure information and historical form information of local blood vessels and surrounding tissues, thereby realizing blood vessel motion tracking. The whole process comprises four main parts of initial extraction of blood vessels, construction of description models, estimation of motion trajectories and updating of the description models. Specifically, the present invention is realized by the following steps:
step 1: initial vessel extraction.
Reading volume image data when initial t is 0, taking the maximum value of blood vessel image intensity and the lower limit value of the nominal value of the normal range of the thoracic cavity image as the range, carrying out histogram normalized enhancement pretreatment on the original volume image to realize the normalization of CT intensity value of the region of interest, and recording the pretreated image data as I0. Then, starting from any preset point on the blood vessel, gradually positioning the center line position of the blood vessel and the diameter of each section of the blood vessel from the volume image by using a three-dimensional region growing and local geometric moment statistic estimation method, and completing the extraction of the digital model of the vein of the blood vessel of interest.
Step 2: and constructing a blood vessel description model.
Aiming at the time phase image data with initial t being 0I0Firstly, using bifurcation point of blood vessel as boundary, dividing the extracted blood vessel into blood vessel segments, subdividing each blood vessel segment into volume image blocks with equal volume according to its diameter, recording the jth local blood vessel of ith segment as
Figure GDA0003171668260000031
As shown in fig. 2, where t is the current phase; then, the volume intensity histogram feature of each block is calculated
Figure GDA0003171668260000032
As a description of its local appearance, the three-dimensional position of the respective piece corresponding to the center of the blood vessel is recorded
Figure GDA0003171668260000033
And calculating the average value of the spatial distances of the adjacent volume blocks
Figure GDA0003171668260000034
Sum variance
Figure GDA0003171668260000035
As the spatial position description of local blood vessels, adjacent blocks are simultaneously combined into volume block pairs, and the spatial included angle between the adjacent volume block pairs is calculated
Figure GDA0003171668260000036
As a morphological description of local blood vessels. Therefore, the strength, the position and the local form of the initial 0-phase form the whole describing model A of the blood vesseltIs marked as
Figure GDA0003171668260000037
And step 3: and estimating the motion track of the blood vessel.
Giving volume image data of the current time phase t, and integrally describing a model and parameters A of a blood vessel of the previous time phaset-1The tracking of the motion trail of the blood vessel is to estimate the position of each block of each segment of the blood vessel in the three-dimensional space under the time phase. For the current time phase, firstly, a volume image is obtained through a histogram normalized image enhancement preprocessing linkData ItThen the previous time phase blood vessel description parameter AtThe determined blood vessel position is an initial quantity, and the total probability P (A) that the blood vessel position is possible is calculated in the current t-phase imaget|It,At-1) The optimal position of the vessel, i.e. the total likelihood P (A)t|It,At-1) To the maximum position. Starting from the blood vessel position at the time phase t-1 as an initial quantity under the current time phase t, and following the log P (A) by a dynamic programming methodt|It,At-1) Iteratively updating the positions of the blocks of each segment by reducing the direction, and calculating the total blood vessel probability P (A) of the iteratively new positionst|It,At-1) Until it reaches a maximum. P (A)t|It,At-1) The specific steps of the calculation comprise:
(1) according to the appearance condition of the local blood vessel, namely the intensity value distribution of the local volume blocks, calculating the appearance possibility of the blood vessel possibly appearing at each block position of each segment of the current blood vessel
Figure GDA0003171668260000041
Wherein the content of the first and second substances,
Figure GDA0003171668260000042
representing the intensity histogram feature of the jth block of the ith segment in the previous time phase;
Figure GDA0003171668260000043
representing the intensity histogram feature of the ith segment jth block at the current position. Chi-square distance is used for intensity change rule of volume blocks between adjacent time phases2(-) description. Furthermore, the position is based on the appearance possibility of each blood vessel in each segment
Figure GDA0003171668260000044
Calculating the probability of the vessel of interest as a whole as a combination of the probabilities of the segments, i.e.
Figure GDA0003171668260000045
(2) According to the spatial position relation of the adjacent volume blocks
Figure GDA0003171668260000046
And
Figure GDA0003171668260000047
parameter calculation of the position probability of the blood vessel possibly appearing at each block position of the current section of the blood vessel
Figure GDA0003171668260000048
The spatial distance distribution of adjacent volume blocks is described by normal distribution N (-) and the two parameters are respectively the mean values calculated by the distance of the corresponding volume block at the previous time
Figure GDA0003171668260000049
Sum variance
Figure GDA00031716682600000410
Figure GDA00031716682600000411
And
Figure GDA00031716682600000412
the three-dimensional positions of the j < th > block of the ith segment and the j +1 < th > block of the ith segment. Furthermore, the possibility of the blood vessel moving to the position is calculated with respect to the position of the blood vessel in the previous time phase, that is, the possibility of the change in the position of the whole blood vessel is determined in common according to the change in the position of each block of each segment of the blood vessel
Figure GDA00031716682600000413
(3) According to the shape condition of local blood vessel, namely the chain space structure parameter condition of blood vessel, calculating the shape possibility that the included angle between adjacent volume block pairs may appear
Figure GDA00031716682600000414
The jth volume block pair of the ith segment in the time of t is
Figure GDA00031716682600000415
And the j +1 th volume block pair is
Figure GDA00031716682600000416
Amount of included angle
Figure GDA00031716682600000417
Representing the amount of volume block diagonal as shown in figure 3. The angle measure is modeled using von mises distribution (Mm), wherein
Figure GDA00031716682600000418
Calculating the probability of the whole blood vessel of interest according to the shape probability of each blood vessel of each segment under the shape by taking the included angle value of each adjacent volume block pair in the previous time phase, wherein kappa is the preset value of the angular stiffness parameter, and kappa is more than or equal to 0
Figure GDA0003171668260000051
In summary, the total probability that the current candidate location is the blood vessel of interest is the appearance probability PfPosition possibility PlPossibility of form
Figure GDA0003171668260000052
Accumulation of the three, i.e.
Figure GDA0003171668260000053
And 4, step 4: and updating the parameters of the blood vessel description model.
To adapt to the complexity of the vessel morphology, the overall description model A is updated phase by phasetIncluding intensity distribution parameters of the whole blood vessel
Figure GDA0003171668260000054
Mean value of spatial position parameters of adjacent volume blocks
Figure GDA0003171668260000055
And standard deviation of
Figure GDA0003171668260000056
Chain angle parameter with blood vessel
Figure GDA0003171668260000057
Set xtA set of model parameters is described for these four vessels. The integral updating operation adopts a preset forgetting factor r which belongs to [0.05,0.3 ]]The method of (1) fusing the last time phase parameter and the current time phase positioned blood vessel vein parameter value by taking the parameter as a proportion, and calculating a model updating value by a recursive filter
Figure GDA0003171668260000058
And (4) processing each time phase volume image data according to the step 3 and the step 4 in an iterative way until all time phase data are processed, and obtaining the time phase-by-time estimation of the blood vessel motion track.
The invention has the characteristics that:
firstly, the invention constructs a vascular morphology modeling mode combining layering and chain structure, and realizes the parametric representation of the vascular morphology structure in three-dimensional space; secondly, aiming at the complex motion situation of the blood vessel in the three-dimensional space, a visual description model of the whole blood vessel is established from appearance expression, spatial relation and local structure form of the local blood vessel, and the problem of integral model description of the blood vessel in the volume image is solved; finally, the invention establishes a dynamic deduction method of the state of the blood vessel model, and provides a set of blood vessel positioning integral scheme from three-dimensional image blood vessel description to four-dimensional image blood vessel motion tracking. The invention constructs a refined blood vessel motion description model, can perform robust and accurate three-dimensional motion tracking on interested blood vessels section by section in the whole respiratory or cardiac cycle, has a tracking result independent of the performance of complex three-dimensional registration operation in the traditional technology, and has efficient and simple integral calculation process.
Description of the drawings:
FIG. 1, general flow diagram of the inventive method;
FIG. 2, a schematic diagram of vessel modeling;
FIG. 3 is a schematic diagram of the amount of included angle between adjacent pairs of volume blocks;
FIG. 4, detailed description;
FIG. 5 is a graph showing the results of the experiment of the present invention:
FIG. (a) shows the initial 0% phase extraction of the left anterior descending coronary artery;
graph (b) is the 20% phase left anterior descending tracking of coronary arteries;
panel (c) shows the 30% phase left anterior descending coronary artery tracking results.
The specific implementation mode is as follows:
the following detailed description of embodiments of the invention is provided in conjunction with the accompanying drawings:
a coronary artery blood vessel motion tracking based on four-dimensional CT images is characterized in that for 10-30 CT image samples which are uniformly collected in time in each respiration or heartbeat cycle, the whole processing flow chart is shown in an attached drawing 1, and the specific details are shown in a drawing 4.
1 extracting blood vessels from the volume image.
1.1 reading volume image data of a time phase with initial t being 0, taking the maximum value of the intensity of a blood vessel image and the lower limit value 240 of the normal range of a thoracic cavity CT image as ranges, carrying out histogram normalized enhancement pretreatment on the original volume image to realize the normalization of the CT intensity value of an interested area, and marking the pretreated image as I0；
1.2 starting from a certain preset point, positioning the central line of the blood vessel in the volume image by a three-dimensional region growing method and local geometric moment statistic estimation, estimating the diameter of the blood vessel along the central line, and obtaining the venation of the blood vessel of interest.
2, constructing a blood vessel description model.
2.1 As shown in FIG. 2, for the initial t-0 phase image I0Dividing the extracted blood vessel into segments by using the bifurcation point of the blood vessel as a boundary, dividing the segments into volume blocks by taking the average diameter n of the current blood vessel segment as 2 times, and marking the jth block of the ith segment as a volume block
Figure GDA0003171668260000061
2.2 calculating the volume intensity histogram feature of each block, dividing the intensity histogram feature into 32 levels, and recording the time phase of the internal block
Figure GDA0003171668260000062
Is characterized by an intensity histogram of
Figure GDA0003171668260000063
2.3 recording the center position of each blood vessel block, time phase internal block
Figure GDA0003171668260000064
In the position of
Figure GDA0003171668260000071
Wherein x, y and z are the axes of the volume image imaging coordinate system;
2.4 calculate the average value of the spatial distance between adjacent volume blocks at t, and the average value of the spatial distance of the i-th section is recorded as
Figure GDA00031716682600000712
And calculating the variance as
Figure GDA0003171668260000072
2.5 adjacent blocks form a volume block pair, the included angle between the adjacent volume block pairs is calculated, and the jth volume block pair of the ith segment is recorded as
Figure GDA0003171668260000073
And the j +1 th volume block pair is
Figure GDA0003171668260000074
Calculating the amount of included angle between lines defined by two volume pairs
Figure GDA0003171668260000075
As shown in fig. 3;
2.6 Forming an Overall descriptive model A of the vessel of interest at the t-phasetI.e. by
Figure GDA0003171668260000076
And 3, tracking the motion trail of the blood vessel.
3.1 given the volume image of the current t-phase and the t-1 phase vascular description model At-1The tracking attention of the blood vessel motion track is to estimate the central three-dimensional position of each block of each segment of the blood vessel under the time phase
Figure GDA0003171668260000077
Preprocessing the t-phase volume image by histogram normalization to obtain a volume image It；
3.2 model A is described as t-1 phase vesselt-1For the initial quantity, it is calculated at ItIs the total probability P (A) of the blood vesselt|It,At-1) The method comprises the following specific steps:
3.2.1 according to the distribution situation of the intensity values, calculating the appearance possibility of the blood vessels possibly appearing at the positions of each block of each current segment
Figure GDA0003171668260000078
K is the number of vessel segments, and the left anterior descending branch of the coronary artery is divided into 2 segments, namely K is 2. k is a radical ofiIs the total number of volume blocks divided equally by 2 times the average diameter n of the current vessel segment. Wherein the likelihood of appearance of each volume being a blood vessel is
Figure GDA0003171668260000079
In the calculation of the formula,
Figure GDA00031716682600000710
representing the intensity histogram feature of the jth block of the ith segment in the previous time phase;
Figure GDA00031716682600000711
representing the intensity histogram feature of the ith segment jth block at the current position. Chi-square distance chi for intensity change rule of volume blocks between time phases2(-) description;
3.2.2 according to the space position relation condition of the adjacent volume blocks, calculating the position possibility of the blood vessel possibly appearing at the position of each block of each current segment
Figure GDA0003171668260000081
Wherein the position probability that each pair of adjacent volume blocks is a blood vessel is
Figure GDA0003171668260000082
In the formula, the spatial distance distribution of adjacent volume blocks is described by normal distribution N (-) and the distance of the corresponding volume block in the previous time is calculated as the mean value
Figure GDA0003171668260000083
Sum variance
Figure GDA0003171668260000084
As a probability calculation parameter;
3.2.3 from the case of the chain spatial relationship of the vessels, the jth volume block pair of the ith segment is
Figure GDA0003171668260000085
And the j +1 th volume block pair is
Figure GDA0003171668260000086
Computing adjacencyOverall shape possibility of included angle between volume block pair
Figure GDA0003171668260000087
Wherein the form possibility of included angle between a certain adjacent volume block pair is
Figure GDA0003171668260000088
The formula adopts von Misses distribution (Mm).)
Figure GDA0003171668260000089
Angle of the mold
Figure GDA00031716682600000810
A possible distribution over time, wherein
Figure GDA00031716682600000811
Represents the relative included angle quantity of the compatible blocks at the current time,
Figure GDA00031716682600000812
representing the diagonal angle quantity of the previous compatible block, wherein kappa is an angular stiffness parameter and is taken as kappa-pi/3;
3.2.4 from three aspects of appearance, spatial relationship and local structure and shape of the local blood vessel, calculating the total possibility that the current candidate position is the interested blood vessel:
Figure GDA00031716682600000813
3.2.5 evaluation-log P (A)t|It,At-1) If the minimum value is reached, otherwise, iteratively calculating 3.2.1-3.2.5 steps according to a dynamic programming method;
3.3 when-log P (A)t|It,At-1) The maximum number of iterations (set to 50) or the variation of the objective function is less than 10-4Stopping iteration, and calculating the optimal position of each block of each segment of the blood vessel at the time phase t
Figure GDA0003171668260000091
4 vessel description model parameter updating
4.1 updating the intensity distribution parameters of the blood vessels
Figure GDA0003171668260000092
By means of recursive filters
Figure GDA0003171668260000093
Calculating an updated value of the model, wherein r is a model forgetting factor, and is taken as 0.1;
4.2 update the spatial position parameters of the neighboring volume blocks, using the same model forgetting factor r, by means of a recursive filter
Figure GDA0003171668260000094
Figure GDA0003171668260000095
4.3 updating the parameters of the chain-type spatial relationship of the blood vessel, using the same model forgetting factor r, calculating by means of a recursive filter
Figure GDA0003171668260000096
And 5, reading the next time phase volume image data, and repeatedly executing the steps 3-4 until all the time phase data are processed.

Claims (1)

1. A vessel motion tracking method based on four-dimensional medical images is characterized by comprising the following steps:
step 1: extracting an initial blood vessel;
reading volume image data when initial t is 0, taking the maximum value of blood vessel image intensity and the lower limit value of the nominal value of the normal range of the thoracic cavity image as the range, carrying out histogram normalized enhancement pretreatment on the original volume image to realize the normalization of CT intensity value of the region of interest, and recording the pretreated image data as I0(ii) a Then starting from any preset point on the blood vessel, gradually positioning the center line position of the blood vessel and the diameter of each section of the blood vessel from the volume image by using a three-dimensional region growing and local geometric moment statistic estimation method, and completing the extraction of the digital model of the vein of the blood vessel of interest;
step 2: constructing a blood vessel description model;
aiming at the phase image data I with initial t being 00Firstly, using bifurcation point of blood vessel as boundary, dividing the extracted blood vessel into blood vessel segments, subdividing each blood vessel segment into volume image blocks with equal volume according to its diameter, recording the jth local blood vessel of ith segment as
Figure FDA0003171668250000011
Wherein t is the current time phase; then, the volume intensity histogram feature of each block is calculated
Figure FDA0003171668250000012
As a description of its local appearance, the three-dimensional position of the respective piece corresponding to the center of the blood vessel is recorded
Figure FDA0003171668250000013
And calculating the average value of the spatial distances of the adjacent volume blocks
Figure FDA0003171668250000014
Sum variance
Figure FDA0003171668250000015
As the spatial position description of local blood vessel, at the same time, adjacent blocks are formed into volume block pair, and calculation is carried outThe amount of included angle of the space between adjacent volume block pairs
Figure FDA0003171668250000016
As a morphological description of local blood vessels; therefore, the strength, the position and the local form of the initial 0-phase form the whole describing model A of the blood vesseltIs marked as
Figure FDA0003171668250000017
And step 3: estimating a blood vessel motion track;
given the volume image data of the current time phase t and the integral description model A of the blood vessel of the previous time phaset-1The tracking of the blood vessel motion track is to estimate the position of each block of each section of the blood vessel in the three-dimensional space at the time phase; for the current time phase, firstly, a volume image data I is obtained through a histogram normalized image enhancement preprocessing linktThen model A is described with a previous time phase vesseltThe determined blood vessel position is an initial quantity, and the total probability P (A) that the blood vessel position is possible is calculated in the current t-phase imaget|It,At-1) The optimal position of the vessel, i.e. the total likelihood P (A)t|It,At-1) A position up to a maximum; starting from the blood vessel position at the time phase t-1 as an initial quantity at the current time phase t, and performing dynamic programming along-logP (A)t|It,At-1) Iteratively updating the positions of the blocks of each segment by reducing the direction, and calculating the total blood vessel probability P (A) of the iteratively new positionst|It,At-1) Until it reaches a maximum; p (A)t|It,At-1) The specific steps of the calculation comprise:
(1) according to the appearance condition of the local blood vessel, namely the intensity value distribution of the local volume blocks, calculating the appearance possibility of the blood vessel possibly appearing at each block position of each segment of the current blood vessel
Figure FDA0003171668250000021
Wherein the content of the first and second substances,
Figure FDA0003171668250000022
representing the intensity histogram feature of the jth block of the ith segment in the previous time phase;
Figure FDA0003171668250000023
representing the intensity histogram feature of the ith segment of the jth block at the current position; chi-square distance is used for intensity change rule of volume blocks between adjacent time phases2(-) description; furthermore, the position is based on the appearance possibility of each blood vessel in each segment
Figure FDA0003171668250000024
Calculating the probability of the vessel of interest as a whole as a combination of the probabilities of the segments, i.e.
Figure FDA0003171668250000025
(2) According to the spatial position relation of the adjacent volume blocks
Figure FDA0003171668250000026
And
Figure FDA0003171668250000027
parameter calculation of the position probability of the blood vessel possibly appearing at each block position of the current section of the blood vessel
Figure FDA0003171668250000028
The spatial distance distribution of adjacent volume blocks is described by normal distribution N (-) and the two parameters are respectively the mean values calculated by the distance of the corresponding volume block at the previous time
Figure FDA0003171668250000029
Sum variance
Figure FDA00031716682500000210
Figure FDA00031716682500000211
And
Figure FDA00031716682500000212
the central three-dimensional positions of the jth block and the jth +1 block of the ith section are obtained; furthermore, the possibility of the blood vessel moving to the position is calculated with respect to the position of the blood vessel in the previous time phase, that is, the possibility of the change in the position of the whole blood vessel is determined in common according to the change in the position of each block of each segment of the blood vessel
Figure FDA00031716682500000213
(3) According to the shape condition of local blood vessel, namely the chain space structure parameter condition of blood vessel, calculating the shape possibility that the included angle between adjacent volume block pairs may appear
Figure FDA00031716682500000214
The jth volume block pair of the ith segment in the time of t is
Figure FDA00031716682500000215
And the j +1 th volume block pair is
Figure FDA00031716682500000216
Amount of included angle
Figure FDA00031716682500000217
Representing a volume block diagonal quantity modeled using von mises distribution (Mm), wherein
Figure FDA00031716682500000218
Calculating the probability of the whole blood vessel of interest according to the shape probability of each blood vessel of each segment under the shape by taking the included angle value of each adjacent volume block pair in the previous time phase, wherein kappa is the preset value of the angular stiffness parameter, and kappa is more than or equal to 0
Figure FDA0003171668250000031
In summary, the total probability that the current candidate location is the blood vessel of interest is the appearance probability PfPosition possibility PlPossibility of form
Figure FDA0003171668250000032
Accumulation of the three, i.e.
Figure FDA0003171668250000033
And 4, step 4: updating parameters of the blood vessel description model;
updating the global description model A phase by phasetIncluding intensity distribution parameters of the whole blood vessel
Figure FDA0003171668250000034
Mean value of spatial position parameters of adjacent volume blocks
Figure FDA0003171668250000035
And standard deviation of
Figure FDA0003171668250000036
Chain angle parameter with blood vessel
Figure FDA0003171668250000037
Set xtDescribing a set of model parameters for the four vessels; the integral updating operation adopts a preset forgetting factor r which belongs to [0.05,0.3 ]]The method takes the time phase parameter as a proportion to fuse the last time phase parameter with the positioned blood vessel vein parameter value of the current time phase,computing model update values by recursive filters
Figure FDA0003171668250000038
And (4) processing each time phase volume image data according to the step 3 and the step 4 in an iterative way until all time phase data are processed, and obtaining the time phase-by-time estimation of the blood vessel motion track.
CN201711205773.3A 2017-11-27 2017-11-27 Vascular motion tracking method based on four-dimensional medical image Active CN108010056B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201711205773.3A CN108010056B (en) 2017-11-27 2017-11-27 Vascular motion tracking method based on four-dimensional medical image

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201711205773.3A CN108010056B (en) 2017-11-27 2017-11-27 Vascular motion tracking method based on four-dimensional medical image

Publications (2)

Publication Number Publication Date
CN108010056A CN108010056A (en) 2018-05-08
CN108010056B true CN108010056B (en) 2021-10-15

Family

ID=62053763

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201711205773.3A Active CN108010056B (en) 2017-11-27 2017-11-27 Vascular motion tracking method based on four-dimensional medical image

Country Status (1)

Country Link
CN (1) CN108010056B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109993729B (en) * 2019-03-20 2021-03-12 北京理工大学 Blood vessel tracking method and device
CN113362370B (en) * 2021-08-09 2022-01-11 深圳市速腾聚创科技有限公司 Method, device, medium and terminal for determining motion information of target object

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8503744B2 (en) * 2007-11-19 2013-08-06 Dekel Shlomi Dynamic method and system for representing a three dimensional object navigated from within
CN101515365B (en) * 2009-03-25 2011-05-11 沈阳东软医疗系统有限公司 Method for automatically separating adherent hyaline-vascular type lung nodule in CT image
CN101799935B (en) * 2009-12-31 2011-09-14 华中科技大学 Dynamic three-dimensional reconstruction method of single-arm X-ray angiogram maps
CN101901498B (en) * 2009-12-31 2011-11-30 华中科技大学 Recursion modelling method of human coronary artery
RU2616987C2 (en) * 2011-03-02 2017-04-19 Конинклейке Филипс Н.В. System and method of medical imaging for image representation to support accurate direction of surgical instrument during blood vessel surgery
CN103077550B (en) * 2012-12-10 2016-04-20 华北电力大学（保定） A kind of four dimensional rebuilding method of non-gate ICUS image sequence medium vessels

Also Published As

Publication number Publication date
CN108010056A (en) 2018-05-08

Similar Documents

Publication Publication Date Title
JP6790179B2 (en) Useful results for real-time diagnostics
US10304197B2 (en) Automated centerline extraction method and generation of corresponding analytical expression and use thereof
CN106037710B (en) Synthetic data-driven hemodynamic determination in medical imaging
Worz et al. Segmentation and quantification of human vessels using a 3-D cylindrical intensity model
CN110638438B (en) Method and system for machine learning based assessment of fractional flow reserve
CN102243759B (en) Three-dimensional lung vessel image segmentation method based on geometric deformation model
CN109478327B (en) Method for automatic detection of systemic arteries in Computed Tomography Angiography (CTA) of arbitrary field of view
EP3061015A2 (en) Vascular characteristic determination with correspondence modeling of a vascular tree
CN112837306B (en) Coronary artery disease lesion functional quantitative method based on deep learning and mesopic theory
CN106570856A (en) Common carotid artery intima-media thickness measuring device and method combining level set segmentation and dynamic programming
CN108010056B (en) Vascular motion tracking method based on four-dimensional medical image
CN112767407A (en) CT image kidney tumor segmentation method based on cascade gating 3DUnet model
CN103324934A (en) Blood vessel center line automatic extraction method based on parallel structure detection and clustering
CN109431532A (en) Artery and vena separation method and device and computer installation based on Perfusion Imaging
CN109003283A (en) A kind of aorta outline segmentation based on active shape model
CN110111354A (en) Liver medical image arteriosomes dividing method and system
CN116108774A (en) Fluid-solid coupling numerical simulation method based on interventional aortic valve
Mu et al. Construction of anatomically accurate finite element models of the human hand and a rat kidney
Son et al. Reconstruction of blood vessel model with adventitia from CT and IVUS images for FSI analysis
Mostafa et al. Improved centerline extraction in fully automated coronary ostium localization and centerline extraction framework using deep learning
Anić et al. Improved three-dimensional reconstruction of patient-specific carotid bifurcation using deep learning based segmentation of ultrasound images
CN115984239A (en) Method, system, device and storage medium for extracting central line of cerebral artery blood vessel
Thai et al. Automatic segmentation and implicit surface representation of dynamic cardiac data
Huseyn DEEP LEARNING METHOD FOR MEDICAL IMAGE ANALYSIS
Chen et al. Simultaneous intracranial artery tracing and segmentation from magnetic resonance angiography by joint optimization from multiplanar reformation

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