CN114732435A - Ultrasonic image spot tracking muscle movement displacement parameter detection method - Google Patents

Ultrasonic image spot tracking muscle movement displacement parameter detection method Download PDF

Info

Publication number
CN114732435A
CN114732435A CN202210350114.3A CN202210350114A CN114732435A CN 114732435 A CN114732435 A CN 114732435A CN 202210350114 A CN202210350114 A CN 202210350114A CN 114732435 A CN114732435 A CN 114732435A
Authority
CN
China
Prior art keywords
ultrasonic
image
ultrasound
spot
block
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
CN202210350114.3A
Other languages
Chinese (zh)
Other versions
CN114732435B (en
Inventor
梁洪涛
郭建中
封一晨
朱春媛
高乐
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.)
Shaanxi Normal University
Original Assignee
Shaanxi Normal 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 Shaanxi Normal University filed Critical Shaanxi Normal University
Priority to CN202210350114.3A priority Critical patent/CN114732435B/en
Publication of CN114732435A publication Critical patent/CN114732435A/en
Application granted granted Critical
Publication of CN114732435B publication Critical patent/CN114732435B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B8/00Diagnosis using ultrasonic, sonic or infrasonic waves
    • A61B8/08Detecting organic movements or changes, e.g. tumours, cysts, swellings
    • A61B8/0833Detecting organic movements or changes, e.g. tumours, cysts, swellings involving detecting or locating foreign bodies or organic structures
    • A61B8/085Detecting organic movements or changes, e.g. tumours, cysts, swellings involving detecting or locating foreign bodies or organic structures for locating body or organic structures, e.g. tumours, calculi, blood vessels, nodules
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B8/00Diagnosis using ultrasonic, sonic or infrasonic waves
    • A61B8/52Devices using data or image processing specially adapted for diagnosis using ultrasonic, sonic or infrasonic waves
    • A61B8/5215Devices using data or image processing specially adapted for diagnosis using ultrasonic, sonic or infrasonic waves involving processing of medical diagnostic data
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/0002Inspection of images, e.g. flaw detection
    • G06T7/0012Biomedical image inspection
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/20Analysis of motion
    • G06T7/246Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/20Analysis of motion
    • G06T7/269Analysis of motion using gradient-based methods
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/10Image acquisition modality
    • G06T2207/10132Ultrasound image
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/20Special algorithmic details
    • G06T2207/20092Interactive image processing based on input by user
    • G06T2207/20104Interactive definition of region of interest [ROI]

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • General Health & Medical Sciences (AREA)
  • Pathology (AREA)
  • Molecular Biology (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Public Health (AREA)
  • Veterinary Medicine (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Multimedia (AREA)
  • Quality & Reliability (AREA)
  • Vascular Medicine (AREA)
  • Ultra Sonic Daignosis Equipment (AREA)

Abstract

A muscle movement displacement parameter detection method for ultrasonic image spot tracking comprises initializing a muscle ultrasonic image set, detecting a significant ultrasonic spot of a first frame image, tracking the ultrasonic spot with matched thickness and granularity, removing abnormal movement parameters of the significant ultrasonic spot, and fitting a movement track curve. The method comprises the steps of screening remarkable ultrasonic spots in an initial frame by adopting an image local autocorrelation function, constructing a region of interest (ROI), establishing a hexagonal search template, tracking coarse-grained ultrasonic spots of the ROI by adopting a block matching method, and obtaining an optimal matching block of each frame of image; tracking the details of the remarkable ultrasonic spots by adopting a Lucas-Kanade optical flow method; the method adopts a random sampling consistency method to remove abnormal motion parameters of the continuous interframes of the remarkable ultrasonic spots, corrects tracking result errors, estimates the muscle motion state, realizes the detection of the motion and deformation of the muscle tissue in a large range, and has certain guiding significance for clinical muscle ultrasound.

Description

Ultrasonic image spot tracking muscle movement displacement parameter detection method
Technical Field
The invention relates to the technical field of ultrasonic medical detection, in particular to a method for measuring skeletal muscle movement parameters by adopting ultrasonic spot tracking.
Background
Skeletal muscle accounts for about 30-40% of human body weight and is an important tissue organ for keeping body posture, regulating metabolism and maintaining functions. Skeletal Muscle Damage (MD) is one of the most common muscle diseases, and the causes of the damage are numerous, including direct trauma such as muscle laceration and contusion due to mechanical force, indirect damage due to age or chronic stress, and muscle diseases such as amyotrophic lateral sclerosis, inflammatory myopathy, muscular dystrophy, etc. due to other factors. Under the conditions of overweight damage or regeneration damage, the muscle function is difficult to achieve effective repair, and economic, safe and effective intervention means are sought to accelerate myoblast expansion and differentiation, inhibit fibrosis reaction accompanied by healing and shorten skeletal muscle damage repair time, so that the method is always a hot problem of medical and physiological research and has important significance and clinical value for improving the motor function of an organism.
The skeletal muscle medical ultrasound has the outstanding advantages of no shooting property, real-time property, convenience and the like, so that the skeletal muscle structure and the motion state can be imaged clearly in real time, and the morphological characteristics of muscle thickness, cross sectional area, muscle bundle length, feather angle and the like can be acquired. As the morphological characteristics and the functional state and the mechanical characteristics of the muscle tissue are closely related, a new way is provided for conveniently and rapidly diagnosing the pathological changes of the muscle tissue. The disorder of fiber arrangement sequence, the diffusion degree of fat echo and the fuzzy degree of muscle boundary caused by skeletal muscle injury are difficult to accurately obtain by the existing medical ultrasonic morphological characteristic processing method, and the specific numerical value of muscle characteristics is determined by mostly depending on manual calibration and measurement and combining subjective historical experience. The method is difficult to meet the requirements of objective, accurate and real-time clinical application, and the development of the ultrasonic imaging technology in muscle research is greatly hindered.
Speckle-tracking imaging (STI), which is based on the principle of high-frame-rate imaging and acoustic Speckle matching, is a new medical ultrasound strain technology that has attracted much attention in recent years, and has made a key breakthrough in Speckle-tracking echocardiography (STE) and Blood Speckle-tracking imaging (BSI). On one hand, the STE evaluates the myocardial contraction function by identifying and tracking deformation information of myocardial tissues in different cardiac cycles, can quantitatively detect the longitudinal, radial, circumferential and ventricular torsion motions of the myocardium through full-volume imaging, can also obtain the myocardial space deformation and real-time motion velocity vectors, and is proved to be an effective tool for evaluating the ventricular function of early myocardial diseases. On the other hand, the BSI is proposed and used to detect blood flow velocity and valve abnormality under high-speed impact of blood flow without being limited by the angle of the sound beam, and has potential application value in rejection reaction, such as using the spot tracking technique of the optical flow method to capture the full information and change characteristics of the carotid blood flow field, unlike the doppler imaging technique which can only capture the blood flow motion information parallel to the sound velocity direction in a limited frame frequency range. The STI theory work undoubtedly has important significance on the rejection and application of cardiovascular diseases, but the prior report applies the STI technology to the mechanism analysis of muscular tissue pathological changes, and the deep research is urgently needed for accurately and quickly tracking the muscle ultrasonic spots.
Disclosure of Invention
The technical problem to be solved by the present invention is to overcome the above disadvantages of the prior art, and to provide a method for detecting muscle movement displacement parameters tracked by ultrasound image speckle, which can accurately evaluate and quickly predict.
The technical scheme adopted for solving the technical problems comprises the following steps:
(1) initializing a set of muscle ultrasound images
Obtaining a muscle ultrasonic image sequence F by adopting B ultrasonic, and carrying out ultrasonic image F on the l frame in the Fl(x, y) initializing Standard Gray image cropped to L1 × L2 size, fl(x,y)∈{f1(x,y),f2(x,y),…,fΠ(x, y) }, l ∈ {1,2, …, Π }, where Π is the number of ultrasound image sequences F; l1 and L2 are finite positive integers.
(2) Detecting salient ultrasound blobs in a first frame image
1) Determining a first frame image f according to equation (1)1Local gray-scale variation E (Δ x, Δ y) when the (x, y) window is translated in different directions (Δ x, Δ y):
E(Δx,Δy)=AΔx2+2CΔxΔy+BΔy2 (1)
Figure BDA0003579558270000031
Figure BDA0003579558270000032
C=w(x,y)*fxfy
where w (x, y) is a Gaussian weighting function representing the convolution operation, fxRepresenting the gradient of the image in the x-direction, fyRepresenting the gradient of the image in the y-direction.
2) Setting an ultrasound speckle response function
Setting an ultrasonic speckle response function R (x, y) according to the formula (2):
R(x,y)=detU-δ(trackU)2 (2)
detU=AB-C2
trackU=A+B
in the formula, detU is a determinant of a matrix U, trackU is a trace of the matrix U, and delta is a control parameter.
3) Setting a threshold R0(x, y) determining the first frame image f according to the formula (3)1(x, y) each ultrasound spot satisfies:
R(x,y)>R0(x,y) (3)
is selected as the significant ultrasound spot hk(x, y) arranging according to the size of the response function in a descending order to form a significant ultrasonic spot set H, and a significant ultrasonic spot Hk(x, y) are as follows:
hk(x,y)={h1(x,y),h2(x,y),…,hΔ(x,y)}
where k e {1,2, …, Δ }, Δ represents the number of significant sets of ultrasound spots H.
4) Determination of the ultrasound speckle h with significancek(x, y) -centered template block Tk(x, y), template Block TkDimension of (x, y) M N, M<L1,N<L2 set of template blocks T corresponding to the set of salient ultrasound spots HROIComprises the following steps:
TROI={T1(x,y),T2(x,y),…,TΔ(x,y)}。
(3) coarse and fine particle size matched ultrasonic speckle tracking
1) At the current frame flSetting a search window S (x, y) with the size of mxn in (x, y), wherein l is more than or equal to 2 and less than or equal to pi, and m<L1,n<L2。
2) Dividing the current frame fl(x, y) are non-overlapping subblocks of the size corresponding to the template block Tk(x, y) are the same, and k is more than or equal to 1 and less than or equal to delta.
3) Initialization
Template block Tk(x, y) is positioned on the central pixel point of the search window S (x, y) to form the current block Ck(x, y), the current block Ck(x, y) around the G block is theta, G is 8 neighboring blocks12345678
4) Starting searching by taking the central pixel point of the search window S (x, y) as the starting point of the large hexagonal template, and determining the current block C according to the formula (4)k(x, y) and template block Tk(x, y) match error function SADk(u,v):
Figure BDA0003579558270000041
Where t (x, y) represents the gray level of the reference frame at (x, y), s (x + u-1, y + v-1) represents the gray level of the current frame at (x + u-1, y + v-1), and (u, v) is the relative displacement motion vector of the matched block.
5) In equation (4), the error function SAD is matchedkThe position with the minimum (u, v) is taken as a vertex, and a matching error function is determinedSADk(u, v) search stop threshold β for corresponding vertexC
Figure BDA0003579558270000042
6) In equation (4), the error function SAD is matchedkThe position with the minimum (u, v) is the central point, and the optimal matching error function SAD in the previous step is searched according to the small hexagonal templatek(u, v) 4 vertices of the centroid, and the match error function SAD for 1 centroid and 4 vertices is determinedk(u, v), match error function SADkThe point at which the (u, v) value is the smallest is the best matching block Copt(x,y)。
7) In equation (4), the error function SAD is matchedk(x,y)≤βCThe searching is stopped, and the matching block corresponding to the vertex is the best matching block Copt(x, y), outputting the best matching block Copt(x, y), establishing an optimal matching block CoptOptical flow field of (x, y):
Figure BDA0003579558270000043
Figure BDA0003579558270000044
Figure BDA0003579558270000045
in the formula
Figure BDA0003579558270000051
Represents the best matching block CoptGraphic spatial gradient of (x, y) coordinates (x, y), IxDenotes the gradient component in the x-direction, IyThe gradient component in the y-direction is represented,
Figure BDA0003579558270000052
representing the optical flow field velocity corresponding to the coordinate (x, y)(u, v) and (u)x,uy;vx,vy) Representing translational motion and deformation motion vectors, ItRepresenting the partial differential of the temporal image.
8) Determining a low pass filter smoothed image
The low pass filter/smoothed image is determined as equation (8):
Figure BDA0003579558270000053
where G (x, y; σ) represents a Gaussian kernel with variance σ.
9) Method for constructing multi-scale affine
The multi-scale affine method is constructed according to the formula (9) as follows:
Lxu-(σ2Lxx+L)ux2Lxyuy+Lyv-(σ2Lyy+L)vy2Lxyvx=-Lt (9)
L(·)=I*G(·)is represented by Copt(x, y) convolution with a gaussian kernel, (· x, xx, y, yy, xy; l istRepresenting the pixel difference between the current frame and the next frame.
10) Determining velocity vectors for salient ultrasound blobs
Determining the velocity vector of a significant ultrasound spot according to equation (10)
Figure BDA0003579558270000054
Figure BDA0003579558270000055
Output each frame image flEach ultrasonic spot h of (x, y)kVelocity vector of (x, y)
Figure BDA0003579558270000056
(4) Removing abnormal motion parameters of a significant ultrasound spot
1) From the data set of the k-th significant ultrasound spot
Figure BDA0003579558270000057
To determine the number of randomly sampled samples Qk
Figure BDA0003579558270000058
In the formula, gamma is a selected local interior point ratio, gamma is 0.2-0.4, q is a minimum data volume, q is 0.1-0.2 pi, P is an initialization confidence probability, and P is 0.80-0.95.
2) Randomly extracting a sample comprising
Figure BDA0003579558270000059
Each test sample is recorded as an estimation model corresponding to the sampling
Figure BDA00035795582700000510
3) Using all velocity vectors
Figure BDA00035795582700000511
For the estimation model
Figure BDA00035795582700000512
Performing inspection, and counting the number of inner points according to the error threshold value epsilon
Figure BDA00035795582700000513
The error threshold e takes the value 0.05.
4) According to
Figure BDA0003579558270000061
Test sample and standard deviation σ (Q)k):
Figure BDA0003579558270000062
Recalculating the estimate by the 2) stepModel (model)
Figure BDA0003579558270000063
Counting the number of inner points again according to the step 3)
Figure BDA0003579558270000064
5)QkWhen Q is more than or equal to Q, Q is the maximum sampling frequency, Q is a limited positive integer, and the abnormal point is cut off and removed to obtain the number of the internal points
Figure BDA0003579558270000065
(5) Fitting a motion trajectory curve
Number of inner points
Figure BDA0003579558270000066
And fitting a motion trail curve of the remarkable ultrasonic spot by adopting a cubic spline method.
In step (3) of the present invention, said step 7) is: in equation (4), the error function SAD is matchedkThe position with the minimum (u, v) is the central point, and the best matching block C is outputopt(x, y), establishing an optimal matching block CoptOptical flow field of (x, y):
Figure BDA0003579558270000067
Figure BDA0003579558270000068
Figure BDA0003579558270000069
in the step (3) of the present invention, the step 7) is: in equation (4), the error function SAD is matchedk(x,y)>βCThen, the best matching block C is outputopt(x, y), building the optimal matching block CoptOptical flow field of (x, y):
Figure BDA00035795582700000610
Figure BDA00035795582700000611
Figure BDA00035795582700000612
in the invention (4), the step 1) is: from the data set of the k-th significant ultrasound spot
Figure BDA00035795582700000613
To determine the number of randomly extracted samples Qk
Figure BDA00035795582700000614
In the formula, gamma is a selected local interior point proportion, the best value of gamma is 0.3, q is the minimum data volume, the best value of q is 0.15 pi, P is an initialization confidence probability, and the best value of P is 0.90.
In the invention, the gray level and gradient information of an image are considered, and the image local autocorrelation function is adopted to screen remarkable ultrasonic spots in an initial frame, so that a plurality of regions of interest (ROI) are constructed; establishing a hexagonal search template, and adopting a block matching method to realize coarse-grained ultrasonic spot tracking of ROI and obtain an optimal matching block of each frame image; adopting a fine-grained ultrasonic spot tracking strategy of a Lucas-Kanade optical flow method to realize the detail tracking of the remarkable ultrasonic spots in the optimal matching block of each frame image; the abnormal motion parameters of the ultrasonic spots which are obvious between continuous frames are removed by adopting a random sampling consistency method, the error of a tracking result is corrected, the estimation of the muscle motion state is realized, the detection of the motion and deformation of the muscle tissue in a large range is realized, and the method has certain guiding significance for clinical muscle ultrasound.
Drawings
FIG. 1 is a process flow diagram of example 1 of the present invention.
FIG. 2 is a plot of the transverse velocity vector of an ultrasound spot.
FIG. 3 is a plot of the transverse velocity vector of an ultrasound spot.
FIG. 4 is a graph of ultrasound spot movement distance between successive frames.
Detailed Description
The present invention will be further described with reference to the accompanying drawings and specific examples, but the present invention is not limited to the examples described below.
Example 1
In fig. 1, the method for detecting muscle movement displacement parameters by ultrasound image speckle tracking according to this embodiment includes the following steps:
(1) initializing a set of muscle ultrasound images
Adopting B-ultrasonic to obtain muscle ultrasonic image sequence F, and carrying out ultrasonic image F on the l frame in Fl(x, y) initializing Standard Gray image cropped to L1 × L2 size, fl(x,y)∈{f1(x,y),f2(x,y),…,fΠ(x, y) }, l ∈ {1,2, …, Π }, where Π is the number of ultrasound image sequences F; l1 and L2 are finite positive integers.
(2) Detecting significant ultrasound speckle in a first frame image
1) Determining a first frame image f according to equation (1)1Local gray-scale variation E (Δ x, Δ y) when the (x, y) window is translated in different directions (Δ x, Δ y):
E(Δx,Δy)=AΔx2+2CΔxΔy+BΔy2 (1)
Figure BDA0003579558270000081
Figure BDA0003579558270000082
C=w(x,y)*fxfy
where w (x, y) is a gaussian weighting function, representing a convolution operation,fxrepresenting the gradient of the image in the x-direction, fyRepresenting the gradient of the image in the y-direction.
2) Setting an ultrasound speckle response function
Setting an ultrasonic speckle response function R (x, y) according to the formula (2):
R(x,y)=detU-δ(trackU)2 (2)
detU=AB-C2
trackU=A+B
in the formula, detU is a determinant of a matrix U, trackU is a trace of the matrix U, and delta is a control parameter.
3) Setting a threshold R0(x, y) determining the first frame image f according to the formula (3)1(x, y) each ultrasound spot satisfies:
R(x,y)>R0(x,y) (3)
is selected as the significant ultrasound spot hk(x, y) arranging according to the size of the response function in a descending order to form a significant ultrasonic spot set H, and a significant ultrasonic spot Hk(x, y) is as follows:
hk(x,y)={h1(x,y),h2(x,y),…,hΔ(x,y)}
where k e {1,2, …, Δ }, Δ represents the number of significant sets of ultrasound spots H.
4) Determination of the ultrasound speckle h with significancek(x, y) -centered template block Tk(x, y), template Block TkDimension of (x, y) M N, M<L1,N<L2 significant ultrasound blob set H corresponds to template set TROIComprises the following steps:
TROI={T1(x,y),T2(x,y),…,TΔ(x,y)}。
in the step, a hexagonal search template is established, a block matching method is adopted, coarse-grained ultrasonic spot tracking of a plurality of interested areas is realized, and an optimal matching block of each frame of image is obtained.
(3) Coarse and fine particle size matched ultrasonic speckle tracking
1) In the current frame flSetting a search window S (x, y) with the size of mxn in (x, y), wherein l is more than or equal to 2 and less than or equal to pi, and m<L1,n<L2。
2) Dividing the current frame fl(x, y) are non-overlapping sub-blocks, the sub-block size being equal to the template block Tk(x, y) are the same, and k is more than or equal to 1 and less than or equal to delta.
3) Initialization
Template block Tk(x, y) is positioned on the central pixel point of the search window S (x, y) to form the current block Ck(x, y), the current block Ck(x, y) around the G block is theta, G is 8 neighboring blocks12345678
4) Starting searching by taking the central pixel point of the search window S (x, y) as the starting point of the large hexagonal template, and determining the current block C according to the formula (4)k(x, y) and template block Tk(x, y) match error function SADk(u,v):
Figure BDA0003579558270000091
Where t (x, y) represents the gray level of the reference frame at (x, y), s (x + u-1, y + v-1) represents the gray level of the current frame at (x + u-1, y + v-1), and (u, v) is the relative displacement of the matched blocks.
In the step, a fine-grained ultrasonic speckle tracking strategy of a Lucas-Kanade optical flow method is adopted to realize the detail tracking of the remarkable ultrasonic speckles in the optimal matching block of each frame image.
5) In equation (4), the error function SAD is matchedk(u, v) minimum position is set as vertex, and matching error function SAD is determinedk(u, v) search stop threshold β for corresponding vertexC
Figure BDA0003579558270000092
6) In equation (4), the error function SAD is matchedkThe position with the minimum (u, v) is the central point, and the optimal matching error function SAD in the previous step is searched according to the small hexagonal templatek(u, v) 4 from the centerVertex, determining the matching error function SAD of 1 center point and 4 verticesk(u, v), match error function SADkThe point at which the (u, v) value is the smallest is the best matching block Copt(x,y)。
7) In equation (4), the error function SAD is matchedk(x,y)≤βCThe searching is stopped, and the matching block corresponding to the vertex is the best matching block Copt(x, y), outputting the best matching block Copt(x, y), establishing an optimal matching block CoptOptical flow field of (x, y):
Figure BDA0003579558270000101
Figure BDA0003579558270000102
Figure BDA0003579558270000103
in the formula
Figure BDA0003579558270000104
Represents the best matching block CoptGraphic spatial gradient of (x, y) coordinates (x, y), IxDenotes the gradient component in the x-direction, IyThe gradient component in the y-direction is represented,
Figure BDA0003579558270000105
representing the optical flow field velocities corresponding to coordinates (x, y), (u, v) and (u)x,uy;vx,vy) Representing translational and deformation motion vectors, ItRepresenting the partial differential of the temporal image.
8) Determining a low pass filter smoothed image
The low pass filter/smoothed image is determined as equation (8):
Figure BDA0003579558270000106
where G (x, y; σ) represents a Gaussian kernel with variance σ.
9) Method for constructing multi-scale affine
The multi-scale affine method is constructed according to the formula (9) as follows:
Lxu-(σ2Lxx+L)ux2Lxyuy+Lyv-(σ2Lyy+L)vy2Lxyvx=-Lt (9)
L(·)=I*G(·)is represented by Copt(x, y) convolution with a gaussian kernel, (· x, xx, y, yy, xy; l istRepresenting the pixel difference between the current frame and the next frame.
10) Determining velocity vectors for salient ultrasound blobs
Determining the velocity vector of a significant ultrasound spot according to equation (10)
Figure BDA0003579558270000107
Figure BDA0003579558270000108
Output each frame image flEach ultrasonic spot h of (x, y)kVelocity vector of (x, y)
Figure BDA0003579558270000109
(4) Abnormal motion parameters for removing significant ultrasound speckle
1) From the data set of the k-th significant ultrasound spot
Figure BDA00035795582700001010
To determine the number of randomly extracted samples Qk
Figure BDA0003579558270000111
Wherein γ is a selected local interior point ratio, γ is 0.2 to 0.4, γ is 0.3, q is a minimum data size, q is 0.1 pi to 0.2 pi, q is 0.15 pi, P is an initialization confidence probability, P is 0.80 to 0.95, and P is 0.90.
2) Randomly extracting a sample comprising
Figure BDA0003579558270000112
Each test sample is recorded as an estimation model corresponding to the sampling
Figure BDA0003579558270000113
3) Using all velocity vectors
Figure BDA0003579558270000114
For the estimation model
Figure BDA0003579558270000115
Performing inspection, and counting the number of inner points according to the error threshold e
Figure BDA0003579558270000116
The error threshold epsilon is 0.05.
4) According to
Figure BDA0003579558270000117
Test sample and standard deviation σ (Q)k):
Figure BDA0003579558270000118
Recalculating the estimation model in 2) steps
Figure BDA0003579558270000119
Counting the number of the inner points again according to the step 3)
Figure BDA00035795582700001110
5)QkWhen Q is more than or equal to Q, Q is the maximum sampling times, and Q takes the value ofCutting off finite positive integers, removing abnormal points to obtain the number of inner points
Figure BDA00035795582700001111
In the step, a random sampling consistency method is adopted to remove abnormal motion parameters of the ultrasonic spots which are obvious between continuous frames, correct tracking result errors, realize the estimation of the muscle motion state and realize the detection of the motion and deformation of muscle tissues in a large range.
(5) Fitting a motion trajectory curve
Number of inner points
Figure BDA00035795582700001112
And fitting a motion track curve of the remarkable ultrasonic spot by adopting a cubic spline method, wherein the transverse velocity vector of the ultrasonic spot is shown in figure 2, the transverse velocity vector of the ultrasonic spot is shown in figure 3, and the moving distance of the ultrasonic spot between successive reading frames is shown in figure 4.
Example 2
The method for detecting the muscle movement displacement parameters tracked by the ultrasonic image spots comprises the following steps:
(1) initializing muscle ultrasound image sets
This procedure is the same as in example 1.
(2) Detecting significant ultrasound speckle in a first frame image
This procedure is the same as in example 1.
(3) Coarse and fine particle size matched ultrasonic speckle tracking
This procedure is the same as in example 1.
(4) Removing abnormal motion parameters of a significant ultrasound spot
1) From the data set of the k-th significant ultrasound spot
Figure BDA0003579558270000121
To determine the number of randomly extracted samples Qk
Figure BDA0003579558270000122
In the formula, gamma is a selected local interior point proportion, gamma is 0.2-0.4, gamma is 0.2, q is a minimum data volume, q is 0.1-0.2 pi, q is 0.1 pi, P is an initialization confidence probability, P is 0.80-0.95, and P is 0.80.
The other steps of this step are the same as in example 1.
(5) Fitting a motion profile
Number of inner points
Figure BDA0003579558270000123
And fitting a motion trail curve of the remarkable ultrasonic spot by adopting a cubic spline method.
Example 3
The method for detecting the muscle movement displacement parameters tracked by the ultrasonic image spots comprises the following steps:
(1) initializing a set of muscle ultrasound images
This procedure is the same as in example 1.
(2) Detecting significant ultrasound speckle in a first frame image
This procedure is the same as in example 1.
(3) Coarse and fine particle size matched ultrasonic speckle tracking
This procedure is the same as in example 1.
(4) Removing abnormal motion parameters of a significant ultrasound spot
1) From the data set of the k-th significant ultrasound spot
Figure BDA0003579558270000124
To determine the number of randomly extracted samples Qk
Figure BDA0003579558270000131
In the formula, gamma is a selected local interior point proportion, gamma is 0.2-0.4, gamma is 0.4, q is a minimum data volume, q is 0.1-0.2 pi, q is 0.2 pi, P is an initialization confidence probability, P is 0.80-0.95, and P is 0.95.
The other steps of this step are the same as in example 1.
(5) Fitting a motion trajectory curve
Number of inner points
Figure BDA0003579558270000132
And fitting a motion trail curve of the remarkable ultrasonic spot by adopting a cubic spline method.
Example 4
The method for detecting the muscle movement displacement parameters tracked by the ultrasonic image spots comprises the following steps:
in examples 1 to 3 above, (1) step(s) and (2) step(s) were the same as in example 1.
(3) Coarse and fine particle size matched ultrasonic speckle tracking
1) About 6) is the same as in example 1.
7) In the formula (4) of the step 4) of (3), the error function SAD is matchedkThe position with the minimum (u, v) is the central point, and the best matching block C is outputopt(x, y), establishing an optimal matching block CoptOptical flow field of (x, y):
Figure BDA0003579558270000133
Figure BDA0003579558270000134
Figure BDA0003579558270000135
(4) abnormal motion parameters for removing significant ultrasound speckle
This step is the same as the corresponding embodiment.
(5) Fitting a motion trajectory curve
Number of inner points
Figure BDA0003579558270000136
And fitting a motion trail curve of the remarkable ultrasonic spot by adopting a cubic spline method.
Example 5
The method for detecting the muscle movement displacement parameters tracked by the ultrasonic image spots comprises the following steps:
in examples 1 to 3 above, (1) step(s) and (2) step(s) were the same as in example 1.
(3) Coarse and fine particle size matched ultrasonic speckle tracking
1) About 6) is the same as in example 1.
7) In the formula (4) of the step 4) of (3), the error function SAD is matchedk(x,y)>βCThen, the best matching block C is outputopt(x, y), establishing an optimal matching block CoptOptical flow field of (x, y):
Figure BDA0003579558270000141
Figure BDA0003579558270000142
Figure BDA0003579558270000143
(4) abnormal motion parameters for removing significant ultrasound speckle
This step is the same as the corresponding embodiment.
(5) Fitting a motion trajectory curve
Number of inner points
Figure BDA0003579558270000144
And fitting a motion trail curve of the remarkable ultrasonic spot by adopting a cubic spline method.
In order to verify the beneficial effects of the invention, the inventor adopts the method of the embodiment of the invention, a full search block matching method and an LK-optical flow method to carry out a comparison experiment, compares the peak signal-to-noise ratio (PSNR) of the current frame to measure the search accuracy, calculates the complexity of the number of all search points of the motion vector to reflect time, and counts the tracking accuracy and the calculation time accuracy. The results are shown in Table 1.
TABLE 1 comparative experimental results of the three methods
Method Average PSNR (dB) Average search point (number) Tracking accuracy (%) Calculating time(s)
Full search block matching 39.69 562 91.21 35.21
LK-light flow method 39.67 456 91.36 34.89
The method of the invention 39.71 196 92.53 23.65
Compared with the PSNR of a muscle ultrasonic image test sequence, the method of the invention has high full search block matching and LK optical flow methods by taking the mean value of all data sets, and shows that the optimal matching block in each frame has the highest measurement search precision; according to the invention, a coarse-fine granularity coupled ultrasonic spot tracking method is adopted, and the average search point is greatly reduced by 183.9% and 132% compared with the matching point which is independently used by a block matching method or an optical flow method; in the aspects of accuracy and calculation time, the method adopts a random sampling consistency strategy to eliminate abnormal data, can ensure the optimal spot motion vector information, and reduces the spot tracking calculation scale. The method can effectively extract the remarkable ultrasonic spots, realizes high-precision tracking of the remarkable ultrasonic spots by a tracking method of coupling of the fineness and the granularity, can eliminate abnormal data, ensures the optimal spot motion vector information, and reduces the spot tracking calculation scale.

Claims (4)

1. A muscle movement displacement parameter detection method for ultrasonic image spot tracking is characterized by comprising the following steps:
(1) initializing a set of muscle ultrasound images
Obtaining a muscle ultrasonic image sequence F by adopting B ultrasonic, and carrying out ultrasonic image F on the l frame in the Fl(x, y) initializing Standard Gray image cropped to L1 × L2 size, fl(x,y)∈{f1(x,y),f2(x,y),…,fΠ(x, y) }, l ∈ {1,2, …, Π }, where Π is the number of ultrasound image sequences F; l1, L2 are finite positive integers;
(2) detecting significant ultrasound speckle in a first frame image
1) Determining a first frame image f according to equation (1)1Local gray-scale variation E (Δ x, Δ y) when the (x, y) window is translated in different directions (Δ x, Δ y):
E(Δx,Δy)=AΔx2+2CΔxΔy+BΔy2 (1)
Figure FDA0003579558260000011
Figure FDA0003579558260000012
C=w(x,y)*fxfy
where w (x, y) is a Gaussian weighting function representing the convolution operation, fxRepresenting the gradient of the image in the x-direction, fyRepresenting the gradient of the image in the y-direction;
2) setting an ultrasound speckle response function
Setting an ultrasonic speckle response function R (x, y) according to the formula (2):
R(x,y)=detU-δ(trackU)2 (2)
detU=AB-C2
trackU=A+B
in the formula, detU is a determinant of a matrix U, trackU is a trace of the matrix U, and delta is a control parameter;
3) setting a threshold R0(x, y) determining the first frame image f according to the formula (3)1(x, y) each ultrasound spot satisfies:
R(x,y)>R0(x,y) (3)
is selected as the significant ultrasound spot hk(x, y) arranging according to the size of the response function in a descending order to form a significant ultrasonic spot set H, and a significant ultrasonic spot Hk(x, y) is as follows:
hk(x,y)={h1(x,y),h2(x,y),…,hΔ(x,y)}
where k ∈ {1,2, …, Δ }, Δ denotes the number of significant sets of ultrasound spots H;
4) determination of the ultrasound speckle h with significancek(x, y) -centered template block Tk(x, y), template Block TkDimension of (x, y) M N, M<L1,N<L2 significant ultrasound blob set H corresponds to template set TROIComprises the following steps:
TROI={T1(x,y),T2(x,y),…,TΔ(x,y)};
(3) coarse and fine particle size matched ultrasonic speckle tracking
1) In the current frame flSetting a search window S (x, y) with the size of mxn in (x, y), wherein l is more than or equal to 2 and less than or equal to pi, and m<L1,n<L2;
2) Dividing the current frame fl(x, y) are non-overlapping sub-blocks, the sub-block size being equal to the template block Tk(x, y) are the same, k is more than or equal to 1 and less than or equal to delta;
3) initialization
Template block Tk(x, y) is positioned on the central pixel point of the search window S (x, y) to form the current block Ck(x, y), the current block Ck(x, y) around the G block is theta, G is 8 neighboring blocks12345678
4) Starting searching by taking the central pixel point of the search window S (x, y) as the starting point of the large hexagonal template, and determining the current block C according to the formula (4)k(x, y) and template block Tk(x, y) match error function SADk(u,v):
Figure FDA0003579558260000021
Wherein t (x, y) represents the gray value of the reference frame at (x, y), s (x + u-1, y + v-1) represents the gray value of the current frame at (x + u-1, y + v-1), and (u, v) is the relative displacement motion vector of the matching block;
5) in equation (4), the error function SAD is matchedkThe position with the minimum (u, v) is taken as a vertex, and a matching error function SAD is determinedk(u, v) search stop threshold β for corresponding vertexC
Figure FDA0003579558260000031
6) In equation (4), the error function SA is matchedDkThe position with the minimum (u, v) is the central point, and the optimal matching error function SAD in the previous step is searched according to the small hexagonal templatek(u, v) 4 vertices of the centroid, and the match error function SAD for 1 centroid and 4 vertices is determinedk(u, v), match error function SADkThe point at which the (u, v) value is the smallest is the best matching block Copt(x,y);
7) In equation (4), the error function SAD is matchedk(x,y)≤βCThe searching is stopped, and the matching block corresponding to the vertex is the best matching block Copt(x, y), outputting the best matching block Copt(x, y), establishing an optimal matching block CoptOptical flow field of (x, y):
Figure FDA0003579558260000032
Figure FDA0003579558260000033
Figure FDA0003579558260000034
in the formula
Figure FDA0003579558260000035
Represents the best matching block CoptGraphic spatial gradient of (x, y) coordinates (x, y), IxDenotes the gradient component in the x-direction, IyRepresenting the gradient component in the y-direction,
Figure FDA0003579558260000036
representing the optical flow field velocities corresponding to coordinates (x, y), (u, v) and (u)x,uy;vx,vy) Representing translational and deformation motion vectors, ItA partial differential representing a temporal image;
8) determining a low pass filter smoothed image
The low pass filter/smoothed image is determined as equation (8):
Figure FDA0003579558260000037
wherein G (x, y; sigma) represents a Gaussian kernel with variance sigma;
9) method for constructing multi-scale affine
The multi-scale affine method is constructed according to the formula (9) as follows:
Lxu-(σ2Lxx+L)ux2Lxyuy+Lyv-(σ2Lyy+L)vy2Lxyvx=-Lt (9)
L(·)=I*G(·)is represented by Copt(x, y) convolution with a gaussian kernel, (· x, xx, y, yy, xy; l istRepresenting the pixel difference between the current frame and the next frame;
10) determining velocity vectors for salient ultrasound blobs
Determining the velocity vector of a significant ultrasound blob according to equation (10)
Figure FDA0003579558260000041
Figure FDA0003579558260000042
Output each frame image flEach ultrasonic spot h of (x, y)kVelocity vector of (x, y)
Figure FDA0003579558260000043
(4) Abnormal motion parameters for removing significant ultrasound speckle
1) From the data set of the k-th significant ultrasound spot
Figure FDA0003579558260000044
In determining random drawingNumber of sampling Qk
Figure FDA0003579558260000045
Wherein gamma is a selected local interior point proportion, gamma is 0.2-0.4, q is the minimum data volume, q is 0.1-0.2 pi, P is an initialization confidence probability, and P is 0.80-0.95;
2) randomly extracting a sample comprising
Figure FDA0003579558260000046
Each test sample is recorded as an estimation model corresponding to the sampling
Figure FDA0003579558260000047
3) Using all velocity vectors
Figure FDA0003579558260000048
For the estimation model
Figure FDA0003579558260000049
Performing inspection, and counting the number of inner points according to the error threshold e
Figure FDA00035795582600000410
The error threshold e takes the value of 0.05;
4) according to
Figure FDA00035795582600000411
Test sample and standard deviation σ (Q)k):
Figure FDA00035795582600000412
Recalculating the estimation model in step 2)
Figure FDA00035795582600000413
Counting the number of the inner points again according to the step 3)
Figure FDA00035795582600000414
5)QkWhen Q is more than or equal to Q, Q is the maximum sampling frequency, Q is a limited positive integer, and the abnormal point is cut off and removed to obtain the number of the internal points
Figure FDA00035795582600000415
(5) Fitting a motion trajectory curve
Number of inner points
Figure FDA00035795582600000416
And fitting a motion trail curve of the remarkable ultrasonic spot by adopting a cubic spline method.
2. The method for detecting muscle movement displacement parameters of ultrasonic image speckle tracking according to claim 1, wherein in the step (3), the step 7) is: in equation (4), the error function SAD is matchedkThe position with the minimum (u, v) is the central point, and the best matching block C is outputopt(x, y), establishing an optimal matching block CoptOptical flow field of (x, y):
Figure FDA00035795582600000417
Figure FDA0003579558260000051
Figure FDA0003579558260000052
3. the ultrasonic image spot tracker according to claim 1The method for detecting the parameters of the tracked muscle movement displacement is characterized in that in the step (3), the step 7) is as follows: in equation (4), the error function SAD is matchedk(x,y)>βCThen, the best matching block C is outputopt(x, y), establishing an optimal matching block CoptOptical flow field of (x, y):
Figure FDA0003579558260000053
Figure FDA0003579558260000054
Figure FDA0003579558260000055
4. the method for detecting muscle movement displacement parameters of ultrasonic image speckle tracking according to claim 1, wherein in (4), the step 1) comprises: from the data set of the k-th significant ultrasound spot
Figure FDA0003579558260000057
To determine the number of randomly extracted samples Qk
Figure FDA0003579558260000056
In the formula, gamma is a selected local interior point proportion, gamma is 0.3, q is the minimum data volume, q takes a value of 0.15 pi, P is an initialization confidence probability, and P takes a value of 0.90.
CN202210350114.3A 2022-04-02 2022-04-02 Muscle movement displacement parameter detection method for ultrasonic image spot tracking Active CN114732435B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210350114.3A CN114732435B (en) 2022-04-02 2022-04-02 Muscle movement displacement parameter detection method for ultrasonic image spot tracking

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210350114.3A CN114732435B (en) 2022-04-02 2022-04-02 Muscle movement displacement parameter detection method for ultrasonic image spot tracking

Publications (2)

Publication Number Publication Date
CN114732435A true CN114732435A (en) 2022-07-12
CN114732435B CN114732435B (en) 2024-05-28

Family

ID=82280008

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210350114.3A Active CN114732435B (en) 2022-04-02 2022-04-02 Muscle movement displacement parameter detection method for ultrasonic image spot tracking

Country Status (1)

Country Link
CN (1) CN114732435B (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101617947A (en) * 2009-07-17 2010-01-06 重庆医科大学 Cardiac muscle texture automatic analyzing method based on spot tracking technology
CN103584919A (en) * 2013-11-12 2014-02-19 中国科学院深圳先进技术研究院 System and method for synchronously detecting multi-modal biological signals
US9629615B1 (en) * 2013-09-06 2017-04-25 University Of Louisville Research Foundation, Inc. Combined B-mode / tissue doppler approach for improved cardiac motion estimation in echocardiographic images

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101617947A (en) * 2009-07-17 2010-01-06 重庆医科大学 Cardiac muscle texture automatic analyzing method based on spot tracking technology
US9629615B1 (en) * 2013-09-06 2017-04-25 University Of Louisville Research Foundation, Inc. Combined B-mode / tissue doppler approach for improved cardiac motion estimation in echocardiographic images
CN103584919A (en) * 2013-11-12 2014-02-19 中国科学院深圳先进技术研究院 System and method for synchronously detecting multi-modal biological signals

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
刘美快;陈斌;: "定量超声评估肌肉功能研究进展", 中国介入影像与治疗学, no. 12, 10 December 2018 (2018-12-10) *

Also Published As

Publication number Publication date
CN114732435B (en) 2024-05-28

Similar Documents

Publication Publication Date Title
Zhao et al. Robust segmentation of intima–media borders with different morphologies and dynamics during the cardiac cycle
US7555151B2 (en) System and method for tracking anatomical structures in three dimensional images
CN110197713B (en) Medical image processing method, device, equipment and medium
Mazaheri et al. Echocardiography image segmentation: A survey
Wei et al. Unsupervised convolutional neural network for motion estimation in ultrasound elastography
Chakraborty et al. 2-D myocardial deformation imaging based on RF-based nonrigid image registration
Snare et al. Real-time scan assistant for echocardiography
CN112581385B (en) Diffusion kurtosis imaging tensor estimation method, medium and device based on multiple prior constraints
CN114119626A (en) Cerebrovascular image segmentation method based on statistical model and multi-scale filtering
Wang et al. Learning-based 3d myocardial motion flowestimation using high frame rate volumetric ultrasound data
CN114732435B (en) Muscle movement displacement parameter detection method for ultrasonic image spot tracking
Chen et al. Competitive Swarm Optimized SVD Clutter Filtering for Ultrafast Power Doppler Imaging
Ghori et al. Echocardiogram analysis using motion profile modeling
Malode New approach of statistical analysis for lung disease diagnosis using microscopy images
Foroughi et al. Elastic registration of 3d ultrasound images
CN101982157A (en) Tracking method of two-dimensional ultrasonic deformation tissue image based on grading center-of-gravity estimation
Mei et al. Toward Improved Accuracy in Quasi-Static Elastography Using Deep Learning.
Asyhari et al. Comparison of Optical Flow Methods: Study About Left Ventricular Tracking in Multi View Echocardiographic Images
CN110927644A (en) Method for measuring mechanical asynchrony in left ventricle based on cardiac nuclear magnetic resonance imaging
Sriraam et al. A Computer-Aided Markov Random Field Segmentation Algorithm for Assessing Fetal Ventricular Chambers
Lee et al. Geometric active model for lesion segmentation on breast ultrasound images
Kim et al. Parametric Imaging with Respiratory Motion Correction for Contrast-Enhanced Ultrasonography
Barbosa et al. 3D motion and strain estimation of the heart: initial clinical findings
Aliniazare et al. Echocardiography frames quantification by empirical mode decomposition method
Wang et al. Echocardiography Segmentation by Fractional Differential and Improved Canny, Analysis by Fourier Descriptor

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