CN106875409A - A kind of light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method - Google Patents
A kind of light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method Download PDFInfo
- Publication number
- CN106875409A CN106875409A CN201710181583.6A CN201710181583A CN106875409A CN 106875409 A CN106875409 A CN 106875409A CN 201710181583 A CN201710181583 A CN 201710181583A CN 106875409 A CN106875409 A CN 106875409A
- Authority
- CN
- China
- Prior art keywords
- region
- sorted
- feature
- sticking patch
- coronal
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 39
- 206010021620 Incisional hernias Diseases 0.000 title claims abstract description 28
- 238000002604 ultrasonography Methods 0.000 title claims abstract description 28
- 238000000605 extraction Methods 0.000 claims abstract description 16
- 238000012545 processing Methods 0.000 claims abstract description 5
- 239000000284 extract Substances 0.000 claims description 11
- 238000001514 detection method Methods 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 10
- 230000007613 environmental effect Effects 0.000 claims description 9
- 239000000523 sample Substances 0.000 claims description 9
- 238000001228 spectrum Methods 0.000 claims description 7
- 230000009466 transformation Effects 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 5
- 230000003595 spectral effect Effects 0.000 claims description 5
- 241001269238 Data Species 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000012217 deletion Methods 0.000 claims description 3
- 230000037430 deletion Effects 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 230000004807 localization Effects 0.000 claims description 3
- 238000011946 reduction process Methods 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 2
- 101150052147 ALLC gene Proteins 0.000 claims 2
- 239000000203 mixture Substances 0.000 claims 1
- 230000008602 contraction Effects 0.000 abstract description 4
- 230000000694 effects Effects 0.000 abstract description 4
- 210000000481 breast Anatomy 0.000 abstract description 3
- 230000002980 postoperative effect Effects 0.000 abstract description 3
- 230000004069 differentiation Effects 0.000 abstract description 2
- 210000001519 tissue Anatomy 0.000 description 10
- 238000012360 testing method Methods 0.000 description 9
- 210000003195 fascia Anatomy 0.000 description 6
- 238000012549 training Methods 0.000 description 6
- 238000002474 experimental method Methods 0.000 description 5
- 210000002784 stomach Anatomy 0.000 description 5
- 238000002513 implantation Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 206010019909 Hernia Diseases 0.000 description 2
- 210000001015 abdomen Anatomy 0.000 description 2
- 210000003484 anatomy Anatomy 0.000 description 2
- 239000002775 capsule Substances 0.000 description 2
- 230000007812 deficiency Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000009499 grossing Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 230000008439 repair process Effects 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 238000000844 transformation Methods 0.000 description 2
- 206010060954 Abdominal Hernia Diseases 0.000 description 1
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 208000035091 Ventral Hernia Diseases 0.000 description 1
- 210000003815 abdominal wall Anatomy 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000009954 braiding Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 238000003748 differential diagnosis Methods 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 210000004907 gland Anatomy 0.000 description 1
- 238000007912 intraperitoneal administration Methods 0.000 description 1
- 210000003205 muscle Anatomy 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000001356 surgical procedure Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- 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
- G06T2207/10132—Ultrasound image
- G06T2207/10136—3D ultrasound image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20036—Morphological image processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20092—Interactive image processing based on input by user
- G06T2207/20104—Interactive definition of region of interest [ROI]
Abstract
The invention belongs to technical field of image processing, specially a kind of light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method.The present invention quantifies to extract from dynamicization three-dimensional breast ultrasound automatically first by textural characteristics extraction algorithm(ABUS)The three-dimensional volume of interest of image(VOI)In region to be sorted associated texture characteristic parameter, for use in the differentiation to sticking patch and manadesma;Then the spatial alternation more sensitive issue such as the postoperative curling of incisional hernia sticking patch, contraction is introduced three-D grain parameter and three-dimensional position parameter to improve the robustness of light-type sticking patch classification and identification algorithm for 2 d texture parameter;Finally feature selecting is carried out using class spacing algorithm and sequential advancement search method.The inventive method feature selecting effect is good, efficiency high, can effectively improve the nicety of grading of light-type incisional hernia sticking patch three-dimensional ultrasound pattern, is easy to automatic Classification and Identification.
Description
Technical field
The invention belongs to technical field of image processing, and in particular to a kind of light-type incisional hernia sticking patch three-dimensional ultrasound pattern is special
Levy extracting method.
Background technology
Generally less than 0.5mm, in more than 3mm, foreign matter residual quantity very little causes its time to light-type sticking patch thickness in braiding aperture
Acoustical signal is weaker, and the echo picture high of the linear in imaging results is not obvious enough.Meanwhile, in ventral hernia repair art, light-type
Sticking patch can be placed on before stomach wall manadesma (Inlay), the space (Sublay) between manadesma and peritonaeum and abdomen between (Onlay), flesh
Peritonaeum inwall (IPOM) four in the chamber position (as shown in Figure 2) related to stomach wall anatomical structure level, and all implantation positions
Put and be all closer to fascia tissue.However, in the cross section view of HHUS (hand-held ultrasound, two-dimentional hand-held ultrasound)
In (cross section and/or sagittal plane), because light-type sticking patch and fascia tissue are generally all shown as a high echogenic area for wire
Domain [9].Therefore, what the linear marking feature of HHUS cross section views was reflected is that light-type sticking patch and fascia tissue are overlapped
Hybrid texture attribute.Manadesma can be regarded as anatomy noise when light-type sticking patch differentiates, not only be harmful to manual detection, more
The linear marking is greatly reduced in computer picture recognition to manadesma and the differential diagnosis value of light-type sticking patch.
Because light-type sticking patch and fascia tissue are overlapped in HHUS images, therefore it is difficult light-type sticking patch from its week
Distinguished in the fascia tissue for enclosing.However, ABUS (automated 3-D breast ultrasound, automatized three-dimensional breast
Gland ultrasound) pass through its coronal-plane (surgical plane) for the discriminating of light-type sticking patch provides extra diagnostic message, new is coronal
Face view can isolate light-type sticking patch from fascia tissue, can so as to present significant light-type sticking patch mesh texture
Depending on change.Therefore, ABUS provides the possibility for selectively analyzing light-type sticking patch coronal-plane textural characteristics, with offer more
Plus accurately characteristic parameter describes the potential of mesh texture, can inherently for the identification of light-type sticking patch provides more accurate
Characteristic parameter.
The content of the invention
It is an object of the invention to propose a kind of light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method, so as to
Light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature can be exactly extracted, automatic classification is known in being easy to subsequent processes
Not.
Light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method proposed by the present invention, first by textural characteristics
Extraction algorithm quantifies region to be sorted in the VOI (volume of interest, volume of interest) for extract ABUS images automatically
Associated texture characteristic parameter, for use in the differentiation to sticking patch and manadesma;Then incisional hernia is mended for 2 d texture parameter
The spatial alternation more sensitive issue such as the postoperative curling of piece, contraction, introduces three-D grain parameter and three-dimensional position parameter to carry
The robustness of light-type sticking patch classification and identification algorithm high;Finally spy is carried out using class spacing algorithm and sequential advancement search method
Levy selection.
It is of the invention to comprise the following steps that:
S1:Build a 2-D for ABUS coronal-planes (two-dimensional, two dimension) foreground mask;
S2:Extract VOI;
S3:Pretreatment operation is carried out to VOI images;
S4:After by pretreatment, one group of region to be sorted of VOI is given, spy is carried out to each region to be sorted one by one
Extraction is levied, 40 features are extracted altogether to each region to be sorted;
S5:40 features to extracting carry out feature selecting, and final selection 11 makes the classification of light-type sticking patch and manadesma
Error reaches the feature of minimum as combinations of features.
The relevant technical details being related to regard to step in the inventive method below are further described specifically:
1. described to build a 2-D foreground mask for coronal-plane in step S1, its step is:
S1.1:0.5 to 0.9 times of ABUS coronal-planes section C of scanning total depth is located at by all1-CnTake out, to all C1-
CnThe pixel of same position does average value processing in image, obtains a coronal-plane average image Cmean;
S1.2:Using Otsu algorithms to image CmeanThreshold process is carried out, bianry image C is obtainedbinary;
S1.3:Using morphology opening operation to CbinaryConnected region in image carries out edge smoothing treatment, and to it most
The black cavity caused by focus or shade in big white connected region is filled, and obtains coronal-plane 2-D foreground mask images
Cmask。
2. in step S2, the extraction VOI, its step is:
S2.1:From ABUS coronal-plane mask images CmaskThe upper left corner starts, and image is divided into the image of 50 × 50 pixels
Block;Run at boundary on the right and lower boundary pixel deficiency cannot piecemeal when, then add appropriate black region;
S2.2:By the coronal-plane position feature of all black picture blocks (the binaryzation value sum of all pixels point is 0)
0 is set to, by all image blocks intersected with foreground mask border (the binaryzation value sum of all pixels point is between 1 to 49)
Coronal-plane position feature be set to 1, by the coronal of all full white image blocks (the binaryzation value sum of all pixels point be 50)
Face position feature is set to 2;
S2.3:One by one by all coronal-plane position features not for 0 image block is chosen for current ROI (region of
Interest, current interest region);Using all ABUS cross sections related to ROI and sagittal plane image-region, by 2-D
ROI expand to the VOI of 3-D;One by one by current VOI send into subsequent characteristics extraction module, until complete to all VOI time
Go through.
3. described to carry out pretreatment operation to VOI images in step S3, its step is:
S3.1:Use 3-D ISRAD (intelligent speckle reducing anisotropic
Diffusion, intelligent spot noise reduction anisotropy parameter) algorithm [1] carries out three-dimensional filtering treatment to the VOI for automatically extracting out,
It is thin without destruction such as light-type sticking patch reticular texture etc. to filter contained speckle noise in ABUS image homogeneous regions as far as possible
Section feature;
S3.2:The single frames cross-sectional image for being pointed to VOI centers using Otsu algorithms [8] does binary conversion treatment, obtains standby
The white connected region of choosing;
S3.3:White connected region using all areas in opening operation deletion bianry image less than 15 pixels;
S3.4:Extract the minimum circumscribed bounding box of each white connected region, calculate the width of bounding box, height and
Apex coordinate;
S3.5:Delete the width of bounding box<The white connected region of 15 pixels;
S3.6:Mark is numbered to all remaining white connected regions, as region to be sorted.
4. in step S4, described to carry out feature extraction to each region to be sorted one by one, its step is:
S4.1:Extract the 2-D textural characteristics in region to be sorted;
S4.2:Extract the 3-D textural characteristics in region to be sorted;
S4.3:The local feature of scan depths residing for region to be sorted is extracted, is based on being swept residing for region to be sorted with one
Retouch the local feature f in the Y- directions of the i.e. ABUS of depthdepthTo characterize the occurrence probability of sticking patch;
S4.4:The environmental characteristic in region to be sorted and hernical sac position relationship is extracted, region to be sorted and hernia are based on one
The location parameter f of capsule position relationshipadjacencyTo characterize the occurrence probability of sticking patch.
Here it is main to use GLCM (gray level co- that are effective by theoretical proof and being used widely
Occurrence matrix, gray level co-occurrence matrixes) gray value in area image to be sorted is converted into texture information by [2].
Meanwhile, for light-type sticking patch the characteristics of coronal-plane can show significant reticular texture, it is also added into being more suited to netted
FD (fractal dimension, fractal dimension) [3] feature of texture analysis.May go out after surgery for light-type sticking patch
The clinical picture of the existing contraction with spatial alternation feature, curling etc., has also been further introduced into 3-D GLCM [4] and 3-D FD
[5] extracting the three-D grain feature in region to be sorted.For ABUS scan attributes and incisional hernia sticking patch implantation position characteristic, also
Propose two position features in region to be sorted (i.e. local feature is corresponding with environmental characteristic).To sum up, 40 ginsengs are used altogether
Count to carry out feature statement to each region to be sorted.Wherein, including 25 2-D textural characteristics, 13 3-D textural characteristics, 1
Based on the local feature of scan depths residing for region to be sorted, 1 special with the environment of hernical sac position relationship based on region to be sorted
Levy.
5. in step S4.1, the 2-D textural characteristics for extracting region to be sorted are carried out in two steps:
The first step, calculates 12 2-D GLCM textural characteristics of region cross section to be sorted single frames section.First, extract
The single frames cross section noise-reduced image at VOI centers;Secondly, upper and lower 5 pixel is done in scan depths direction to each region to be sorted
Region extends;Then, it is partitioned into the cross section single frames section in each region to be sorted;Finally, 12 are calculated to single frames section
Descriptor, as the 2-D GLCM textural characteristics that the region cross section to be sorted single frames is cut into slices;
Second step, calculates 12 2-D GLCM textural characteristics and 1 2-D FD of coronal-plane Slice Sequence in region to be sorted
Feature.First, all coronal-plane Slice Sequences corresponding with each scan depths residing for region to be sorted are extracted;Secondly, it is right
In each descriptor fi(i=1...n, n are the numbers of plies of coronal-plane Slice Sequence), uses descriptor fiSuccessively to every coronal-plane
Section is calculated, and draws one group of characteristic value [f1...fn];Finally, [f is taken1...fn] average FmeanAs this descriptor pair
In the characteristic value of coronal-plane Slice Sequence in region to be sorted.
For 2-D GLCM, such as definition initially by scholar Haralick in document [2] to GLCM, in two dimensional image
Two differences in spatial location of pixel can use motion vectorTo describe.D is the distance between two pixels,
It is two pixels and the angle of reference axis.Set a distance d is given for one, on 4 independent directionsThere may be 8 adjacent pixels altogether to occurring, as shown in Figure 4.So, in two dimensional image
Direction is separated by a pair of pixels pair of d, the probability for occurring with gray scale i and j respectively, i.e.,It is designated as pij.Will be by pijGroup
Into matrix normalization be the gray level co-occurrence matrixes that can obtain image.
Haralick proposes 14 textural characteristics calculated from GLCM in document [2], altogether.The present invention have selected it
In 12 features, comprising energy (f1- Energy), contrast (f2- Contrast), correlation (f3- Correlation), variance
(f4- Variance), homogeney (f5- Homogeneity), average (f6- Sum Average), entropy (f7- Entropy), from phase
Close (f8- Autocorrelation), otherness (f9- Dissimilarity), cluster shade (f10- Cluster Shade), collection
Group's protrusion (f11- Cluster Prominence) and maximum probability (f12-Maximum Probability)。
Wherein, NgBe grey level, p (i, j) is the probability that gray scale occurs to (i, j), i.e. gray level co-occurrence matrixes normalization
Result.The average value and standard deviation of matrix row and column isWith
For 2-D FD, two-dimensional fractal dimension is estimated by the power spectrum of image Fourier transformation.Use following FFT
(fast fourier transform, Fast Fourier Transform (FFT)) carries out DFT (discrete fourier to two dimensional image
Transform, discrete Fourier transform):
Wherein, I is the two dimensional image region that size is (M, N), and u and v is respectively the spatial frequency (u=in x and y directions
0,1 ... M-1, v=0,1 ... N-1).Power spectral density P is estimated as follows by F (u, v):
P (u, v)=| F (u, v) |2 (14)
To calculate 2-D FD, P is carried out averagely along the radial section direction across FFT frequency domains.Frequency space is by equably
It is divided into 24 directions, and 30 points of equably being sampled to the radial component in each direction.Calculate log (Pf) to log's (f)
Least square fitting, whereinRadial frequency [10] is represented, then FD is relevant to this double-log in following form
Slope of a curve β:
Wherein, DTIt is topological dimension, for two dimensional image, DT=2.
6. in step S4.2, the 3-D textural characteristics for extracting region to be sorted, to the volume number in each region to be sorted
According to calculating 12 3-D GLCM textural characteristics and 1 3-D FD feature.
For 3-D GLCM, 2-D GLCM are extended to 3-D GLCM by the method for document [4].Two in 3-D view
The differences in spatial location of tissue points can use motion vectorTo describe.D is the distance between two tissue points,It is two bodies
Azimuth between vegetarian refreshments, θ is the zenith angle between two tissue points.Set a distance d is given for one, on 13 independent directions, altogether
There may be 26 adjacent voxels to occurring, as shown in Figure 5.Here, still selective extraction and identical 12 in 2-D GLCM
3-D GLCM textural characteristics.
For 3-D FD, three-dimensional fractal dimension is estimated by the power spectrum of the 3-D Fourier transformations of volume images.Use
Following 3-D FFT to whole 3-D view carry out 3-D DFT:
Wherein, I be size be (M, N, K) 3-D view region, u, v and w be respectively in the space of x, y and z directionss frequently
Rate.Power spectral density P estimates as follows:
P (u, v, w)=| F (u, v, w) |2 (17)
To calculate 3-D FD, P is carried out averagely along the radial sector direction across 3-D FFT frequency domains.Frequency space is equal
Decile is carried out in 24 azimuth directions and 12 zenith angular direction evenly, and the radial component in each direction is equably sampled
30 points.Calculate log (Pf) to the least square fitting of log (f), whereinRepresent radial frequency
[10], then 3-D FD are relevant to the slope β of this double logarithmic curve in following form:
Wherein, DTIt is topological dimension, for 3-D view, DT=3.
7. in step S4.3, the local feature for extracting scan depths residing for region to be sorted, based on stomach wall along ABUS
The anatomical features and four kinds of sticking patch implantation position features of strain less repair formula of Y-direction are scanned, light-type sticking patch is in belly
Occurrence probability in 1-3cm depth areas below skin is maximum.It is therefore proposed that one is based on scan depths residing for region to be sorted
The local feature f in (the Y- directions of ABUS)depthTo characterize the occurrence probability of sticking patch, and along ABUS scan depths direction to the part
Feature situation as shown in Figure 6 is divided, i.e., carry out interval division to skin depth below from probe interface, respectively interval
0~1cm, 1~3cm of interval and 3~6cm of interval.According to the depth parameter in region to be sorted, to local feature fdepthSet such as
Under:1) 0~1cm of interval, light-type sticking patch occurrence probability is medium on the upper side, i.e. fdepthBe set to 3., reason be this regional location compared with
Shallow, only Onlay and Inlay arts formula is possible to place sticking patch in this region.But both are higher because of recurrence rate, at present should in clinic
With not too much many.2) 1~3cm of interval, light-type sticking patch occurrence probability highest, i.e. fdepthBe set to 4., reason be this region just
It is well the position for placing sticking patch in clinic using most Sublay and IPOM arts formulas at present;3) it is interval in 3~6cm, with depth
The increase of degree, the possibility for placing sticking patch is gradually reduced, therefore from top to bottom, i.e. the f in this regiondepth2. and 1. it is set to.
8. in step S4.4, the environmental characteristic for extracting region to be sorted and hernical sac position relationship is divided to following two stepping
OK.
The first step, to determine hernical sac position, proposes a kind of quick hernical sac detection location algorithm based on ABUS data.It is first
First, all black objects in detection ABUS coronal images;Secondly, after pseudo- black objects are filtered, maximum black is calculated
The size of target i.e. hernical sac in X-, Y- and Z- direction;3rd, to ensure that whole hernical sac is all contained among VOI volumes,
Hernical sac size is all expanded into 40 pixels in three directions;4th, the VOI appearances cut out containing hernical sac are concentrated from ABUS volume datas
Product;5th, spot noise reduction process is carried out to VOI using 3-D ISRAD algorithms;Finally, divide from every frame cross-sectional image of VOI
Hernical sac profile is cut out, detection and localization process of the algorithm to hernical sac in ABUS data is completed.
Second step, after the completion of hernical sac detection, is adjusted to coronal-plane position feature.Pair with hernical sac coronal-plane projected area
The coronal-plane position feature in the region to be sorted that domain overlaps or intersects adds 2;Pair with hernical sac coronal-plane projected area within 3cm
The coronal-plane position feature in region to be sorted adds 1.There is the probability of sticking patch in the to be sorted region nearer apart from hernical sac position
It is bigger, it is therefore proposed that a location parameter f based on region to be sorted with hernical sac position relationshipadjacencyTo characterize sticking patch
Occurrence probability.So, using the coronal-plane position feature value in region to be sorted as environmental characteristic fadjacencyValue.Treat point
The f in class regionadjacencyValue is bigger, and corresponding sticking patch occurrence probability is bigger.
9. in step S5,40 features of described pair of extraction carry out feature selecting, and its step is:
(1) between using DBC (distance between class, class spacing) method [6] to calculate the class of each feature respectively
Away from.First, all 18 cases (case group) and incision tract 55 without sticking patch containing light-type sticking patch in incision tract
In case (control group), with reference to this experiment case load, 278 and the present invention for representing all different cases are artificially chosen by doctor
The area size's identical typical sample region to be sorted for automatically obtaining.Only contain the typical case of light-type sticking patch including 125
Region and 153 representative regions for only containing manadesma;Secondly, 40 features extracted before use are to all 278 sample areas
Do calculation of characteristic parameters and normalization in domain;3rd, the class spacing of each feature is calculated respectively;
(2) being sorted from big to small by class spacing to 40 features, chooses larger preceding 25 features of class spacing, as spy
Levy the primary dcreening operation result of selection;
(3) feature to tentatively select 25 uses SFS (sequential forward selection, before order
Enter search) method [7] selected, to obtain making classification accuracy rate highest combinations of features.First, it is true by doctor from 278
In fixed sample, training set 139 and test set 139 are randomly selected.Secondly, grader selects SVM (support vector
Machine, SVMs), grader is trained with training set, with test set testing classification effect.In order to reduce experiment sample
Limited caused error, carries out 100 random experiments altogether, is evenly dividing training set and test set at random again every time.Root
According to final result, when 4 2-D textural characteristics (energy, homogeney, entropy, the fractal dimension of two-dimensional coronal face image) of selection, 6
3-D textural characteristics (energy, correlation, homogeney, entropy, auto-correlation, the fractal dimension of three-dimensional volume to be sorted) and 1 area to be sorted
Three dimensionality location parameter (the f in domainadjacency) this 11 features when, can reach light-type sticking patch and the error in classification of manadesma
Minimum, so finally choosing 11 features as combinations of features.
As it was previously stated, being extracted 40 features altogether to each region to be sorted.Due to having one between these features
Fixed redundancy, and the classification capacity of each feature is also not quite similar.Meanwhile, in order to reduce intrinsic dimensionality, improve classification accurate
Property, reduce classification the working time.The present invention carries out feature selecting by from DBC and SFS to above-mentioned 40 features.Finally, select
Taking 11 makes the error in classification of light-type sticking patch and manadesma reach the feature of minimum as combinations of features.They are 4 2-D textures
Feature (energy, homogeney, entropy, the fractal dimension of two-dimensional coronal face image), 6 3-D textural characteristics be (three-dimensional volume to be sorted
Energy, correlation, homogeney, entropy, auto-correlation, fractal dimension) and 1 three dimensionality location parameter (f in region to be sortedadjacency)。
Compared with prior art, it is of the invention not only to the light-type sticking patch having a extensive future, using ABUS technologies to it
Effectively reliable feature extraction is carried out, and realizes the optimization choosing to light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature
Select, effectively improve the accuracy of Classification and Identification, reduce the classification working time.The inventive method feature selecting effect is good, efficiency high,
The nicety of grading of light-type incisional hernia sticking patch three-dimensional ultrasound pattern can be effectively improved.
Brief description of the drawings
Fig. 1 is light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method flow chart of the invention.
Fig. 2 is abdominal wall structure and four kinds of sticking patch implantation position schematic diagrames.Light-type sticking patch (black dotted lines) can be positioned over
Before stomach wall manadesma (Onlay), between being positioned over flesh (Inlay), the space (Sublay) between manadesma and peritonaeum is positioned over, or be positioned over
Intraperitoneal peritonaeum inwall (IPOM).
Fig. 3 is the typical ABUS images of light-type sticking patch and manadesma.Wherein, (a) and (b) is respectively light-type sticking patch and muscle
Image of the film in ABUS coronal-planes;C () and (d) is the two image in ABUS cross sections.
Fig. 4 is pixel to the spatial relation figure in two dimensional image.The central pixel point given for one is (white
Color), it is independent at 4On direction, there may be 8 pixels at a distance of normalized cumulant d altogether to (grey).
Fig. 5 is pixel to the spatial relation figure in 3-D view.The center tissue points given for one are (white
Color), it is independent at 13On direction, there may be 26 voxels at a distance of normalized cumulant d altogether to (grey).
Local feature value setting schematic diagrames of the Fig. 6 for the ABUS imaging schematic diagrams in stomach wall region and based on scan depths.
Fig. 7 is automatically extracted for ROI's.Wherein, (a) foreground mask image Cmask;B () is by CmaskAdded at lower boundary black
Color region, and it is divided into the image block of 50 × 50 pixels;C () ROI extracts result, coronal-plane position feature be 1 ROI with thin white
Color frame represents, coronal-plane position feature is that 2 ROI is represented with thick white box.
Fig. 8 is VOI expansion process.Wherein, three orthogonal plane (A of (a) ABUS original volume data:Cross section, S:Arrow
Shape face, C:Coronal-plane);The three orthogonal plane views of (b) VOI;The 3-D view of (c) VOI.
Fig. 9 .VOI image preprocessing process schematics.Wherein, (a) is located at the single frames cross section original image at VOI centers;
(b) 3-D ISRAD noise-reduced images;(c) bianry image;D the small objects of () image (c) delete result;E () image (d) most
Small circumscribed bounding box extraction results;F () narrow goal is deleted and alternative target numbering annotation results.
Figure 10 is that the cross section single frames in 5 regions to be sorted in Fig. 9 (f) is cut into slices and coronal-plane Slice Sequence.Wherein, (a)
5 single frames section in cross section;5 groups of Slice Sequences of (b) coronal-plane.
Figure 11 is that the quick hernical sac based on ABUS data detects location algorithm to automatically extracting result containing hernical sac VOI.(a)
Three orthogonal plane (A of ABUS original volume data:Cross section, S:Sagittal plane, C:Coronal-plane);Three orthogonal planes of (b) VOI
View;The 3-D view of (c) VOI.
Figure 12 is the adjustment result of coronal-plane position feature in region to be sorted.A the three-dimensional reconstruction of () hernical sac testing result is regarded
Figure;B () hernical sac coronal-plane is projected and all regions to be sorted;C () needs to be adjusted coronal-plane position feature to be sorted
Region.The region to be sorted that coronal-plane position feature adds 2 represents with thick white box, the area to be sorted that coronal-plane position feature adds 1
Domain is represented with thin white box.
Figure 13 is whole 40 DBC result of calculations of feature.Wherein feature f39It is scan depths feature fdepth, feature f40
It is region coronal-plane position feature f to be sortedadjacency。
Specific embodiment
The present invention is illustrated with reference to the accompanying drawings and detailed description, it is clear that described embodiment is only
It is a part of embodiment of the invention, rather than whole embodiments.Based on the embodiment in the present invention, ordinary skill people
The every other embodiment that member is obtained under the premise of creative work is not made, belongs to the scope of protection of the invention.
Referring to the drawings 1, the light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method of the technical program include with
Under step:
S1:A 2-D foreground mask for ABUS coronal-planes is built, shown in experimental result such as Fig. 7 (a);
(1) it is located at 0.5 to 0.9 times of ABUS coronal-planes section C of scanning total depth by all1-CnTake out, to all C1-Cn
The pixel of same position does average value processing in image, obtains a coronal-plane average image Cmean;
(2) using Otsu algorithms to image CmeanThreshold process is carried out, bianry image C is obtainedbinary;
(3) using morphology opening operation to CbinaryConnected region in image carries out edge smoothing treatment, and maximum to it
The black cavity caused by focus or shade in white connected region is filled, and obtains coronal-plane 2-D foreground mask images
Cmask。
S2:Extract VOI;
(1) from ABUS coronal-plane mask images CmaskThe upper left corner starts, and image is divided into the image block of 50 × 50 pixels.
Run at boundary on the right and lower boundary pixel deficiency cannot piecemeal when, then add appropriate black region, experimental result such as Fig. 7
Shown in (b);
(2) the coronal-plane position feature of all black picture blocks (the binaryzation value sum of all pixels point is 0) is put
It is 0, by all image blocks intersected with foreground mask border (the binaryzation value sum of all pixels point is between 1 to 49)
Coronal-plane position feature is set to 1, by the coronal-plane of all full white image blocks (the binaryzation value sum of all pixels point is 50)
Position feature is set to 2, shown in experimental result such as Fig. 7 (c);
(3) one by one by all coronal-plane position features not for 0 image block is chosen for current ROI.Using all with ROI phases
The ABUS cross sections of pass and sagittal plane image-region, the ROI of 2-D is expanded to the VOI of 3-D.After current VOI is sent into one by one
Continuous characteristic extracting module, until completing the traversal to all VOI, experimental result is as shown in Figure 8.
S3:Pretreatment operation is carried out to VOI images, experimental result is as shown in Figure 9;
(1) three-dimensional filtering treatment is carried out to the VOI for automatically extracting out using 3-D ISRAD algorithms;
(2) the single frames cross-sectional image for being pointed to VOI centers using Otsu algorithms does binary conversion treatment, obtains alternative white
Connected region;
(3) the white connected region using all areas in opening operation deletion bianry image less than 15 pixels;
(4) the minimum circumscribed bounding box of each white connected region is extracted, width, height and the top of bounding box is calculated
Point coordinates;
(5) width of bounding box is deleted<The white connected region of 15 pixels;
(6) mark is numbered to all remaining white connected regions, as region to be sorted.
S4:After by pretreatment, one group of region to be sorted of VOI is given, each region to be sorted can be carried out one by one
Feature extraction, 40 features are extracted to each region to be sorted altogether;
(1) the 2-D textural characteristics in region to be sorted are extracted;
A () calculates 12 2-D GLCM textural characteristics of region cross section to be sorted single frames section.First, VOI is extracted
The single frames cross section noise-reduced image at center;Secondly, the region of upper and lower 5 pixel is done in scan depths direction to each region to be sorted
Extension;Then, it is partitioned into the cross section single frames section in each region to be sorted;Finally, 12 descriptions are calculated to single frames section
Symbol, as the 2-D GLCM textural characteristics that the region cross section to be sorted single frames is cut into slices.Figure 10 (a) is given as shown in Fig. 9
5 region cross section to be sorted single frames sections;
B () calculates 12 2-D GLCM textural characteristics and 1 2-D FD feature of coronal-plane Slice Sequence in region to be sorted.
First, all coronal-plane Slice Sequences corresponding with each scan depths residing for region to be sorted are extracted;Secondly, for each
Descriptor fi(i=1...n, n are the numbers of plies of coronal-plane Slice Sequence), uses descriptor fiEvery coronal-plane is cut into slices successively into
Row is calculated, and draws one group of characteristic value [f1...fn];Finally, [f is taken1...fn] average FmeanAs this descriptor for treating point
The characteristic value of class region coronal-plane Slice Sequence.As shown in Fig. 95 region coronal-planes to be sorted are given such as Figure 10 (b) to cut
Piece sequence;
12 2-D GLCM textural characteristics include energy (f1- Energy), contrast (f2- Contrast), correlation (f3-
Correlation), variance (f4- Variance), homogeney (f5- Homogeneity), average (f6- Sum Average), entropy
(f7- Entropy), auto-correlation (f8- Autocorrelation), otherness (f9- Dissimilarity), cluster shade (f10-
Cluster Shade), cluster protrude (f11- Cluster Prominence) and maximum probability (f12-Maximum
Probability), calculated by formula 1-12 respectively;
For 2-D FD, two-dimensional fractal dimension is estimated by the power spectrum of image Fourier transformation.Use formula 13
FFT carries out DFT to two dimensional image, and spectrum density P estimated by F (u, v) by formula 14;
To calculate 2-D FD, P is carried out averagely along the radial section direction across FFT frequency domains.Frequency space is by equably
It is divided into 24 directions, and 30 points of equably being sampled to the radial component in each direction.Calculate log (Pf) to log's (f)
Least square fitting, whereinRadial frequency is represented, then FD is relevant to this double logarithmic curve by formula 15
Slope β;
(2) the 3-D textural characteristics in region to be sorted are extracted, 12 3-D GLCM textural characteristics is calculated and 1 3-D FD is special
Levy.2-D GLCM are extended to 3-D GLCM by the method for document [4].For 3-D FD, three-dimensional fractal dimension is by volume images
The power spectrum of 3-D Fourier transformations estimated.3-D is carried out to whole 3-D view using the 3-D FFT of formula 16
DFT, power spectral density P is estimated by formula 17;
To calculate 3-D FD, P is carried out averagely along the radial sector direction across 3-D FFT frequency domains.Frequency space is equal
Decile is carried out in 24 azimuth directions and 12 zenith angular direction evenly, and the radial component in each direction is equably sampled
30 points.Calculate log (Pf) to the least square fitting of log (f), whereinRepresent radial frequency,
Then 3-D FD are relevant to the slope β of this double logarithmic curve by formula 18;
(3) local feature of scan depths residing for region to be sorted is extracted, proposes that one is based on being swept residing for region to be sorted
Retouch the local feature f in depth (the Y- directions of ABUS)depthTo characterize the occurrence probability of sticking patch, and along ABUS scan depths direction pair
Local feature situation as shown in Figure 6 is divided.According to the depth parameter in region to be sorted, to local feature fdepthSet
It is as follows:1) interval 0~1cm, fdepth=3., and 2) interval 1~3cm, fdepth=4., and 3) it is interval in 3~6cm, by this region
fdepth2. and 1. feature is set to;
(4) environmental characteristic in region to be sorted and hernical sac position relationship is extracted, proposes that one is based on region to be sorted and hernia
The location parameter f of capsule position relationshipadjacencyTo characterize the occurrence probability of sticking patch.
To determine hernical sac position, a kind of quick hernical sac detection location algorithm based on ABUS data is proposed.First, detect
All black objects in ABUS coronal images;Secondly, after pseudo- black objects are filtered, maximum black target is calculated also just
It is size of the hernical sac in X-, Y- and Z- direction;3rd, to ensure that whole hernical sac is all contained among VOI volumes, by hernical sac size
All expand 40 pixels in three directions;4th, concentrated from ABUS volume datas and cut out the VOI volumes containing hernical sac;5th, make
Spot noise reduction process is carried out to VOI with 3-D ISRAD algorithms;Finally, it is partitioned into hernical sac wheel from every frame cross-sectional image of VOI
Exterior feature, completes detection and localization process of the algorithm to hernical sac in ABUS data.Algorithm containing hernical sac VOI to automatically extracting result as schemed
Shown in 11.
After the completion of hernical sac detection, coronal-plane position feature is adjusted.It is pair Chong Die with hernical sac coronal-plane view field
Or the coronal-plane position feature in intersecting region to be sorted adds 2;It is pair to be sorted within 3cm with hernical sac coronal-plane projected area
The coronal-plane position feature in region adds 1.It is final using the coronal-plane position feature value in region to be sorted as environmental characteristic
fadjacencyValue.The adjustment result of coronal-plane position feature in region to be sorted is as shown in figure 12.
S5:40 features to extracting carry out feature selecting, and final selection 11 makes the classification of light-type sticking patch and manadesma
Error reaches the feature of minimum as combinations of features;
(1) the class spacing of each feature is calculated respectively using DBC.First, in whole of the incision tract containing light-type sticking patch
In 55 cases (control group) of 18 cases (case group) and incision tract without sticking patch, 278 are selected manually by doctor
The area size's identical typical sample region to be sorted automatically obtained with this paper.Only mended containing light-type including 125
The representative region of piece and 153 representative regions for only containing manadesma.Secondly, 40 features extracted before use are to all 278
Individual sample areas do calculation of characteristic parameters and normalization.3rd, the class spacing of each feature is calculated respectively and 40 features are pressed
Class spacing is ranked up from big to small.Finally, primary dcreening operation result of larger preceding 25 features of class spacing as feature selecting is chosen.
Experimental result is as shown in figure 13;
(2) being sorted from big to small by class spacing to 40 features, chooses larger preceding 25 features of class spacing;
(3) feature to tentatively select 25 is selected using SFS.First, from 278 samples for having been determined by doctor
In this, training set 139 and test set 139 are randomly selected.Secondly, grader selects SVM, and grader is trained with training set,
With test set testing classification effect.In order to reduce the limited caused error of experiment sample, 100 random experiments are carried out altogether, often
It is secondary to be evenly dividing training set and test set at random again.Finally, choosing 11 makes the error in classification of light-type sticking patch and manadesma
The feature of minimum is reached as combinations of features.They be 4 2-D textural characteristics (energy of two-dimensional coronal face image, homogeney,
Entropy, fractal dimension), 6 3-D textural characteristics (energy, correlation, homogeney, entropy, auto-correlation, FRACTAL DIMENSIONs of three-dimensional volume to be sorted
Number) and 1 three dimensionality location parameter (f in region to be sortedadjacency)。
The present invention chooses makes the error in classification of light-type sticking patch and manadesma reach the feature of minimum as combinations of features, realizes
To the optimum choice of light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature, effectively improve the accuracy of Classification and Identification, reduce
The classification working time.The present invention is for 2 d texture parameter to the spatial alternations such as the postoperative curling of incisional hernia sticking patch, contraction sensitivity
Problem, it is proposed that using three-D grain parameter and be aided with ABUS three dimensionalities location parameter and extract light-type incisional hernia sticking patch feature
Method, the nicety of grading of light-type incisional hernia sticking patch three-dimensional ultrasound pattern can be effectively improved.
Bibliography
[1]Wu J,Wang Y,Yu J,Shi X,Zhang J,Chen Y,Pang Y."Intelligent speckle
reducing anisotropic diffusion algorithm for automated 3-D ultrasound
images."Journal of the Optical Society of America A,2015,32(2):248-257.
[2]Haralick R.M.,Shanmugam K.,Dinstein I.H."Textural features for
image classification."IEEE Transactions on Systems,Man and Cybernetics 1973,
SMC-3(6):610-621.
[3]Chaudhuri B.B.,Sarkar N."Texture segmentation using fractal
dimension."IEEE Transactions on Pattern Analysis and Machine Intelligence,
1995,17(1):72-77.
[4]Chen W.,Giger M.L.,Li H.,Bick U.,Newstead G.M."Volumetric texture
analysis of breast lesions on contrast-enhanced magnetic resonance images."
Magnetic Resonance in Medicine,2007,58(3):562-571.
[5]Tang D.,Marangoni A.G."3D fractal dimension of fat crystal
networks."Chemical Physics Letters,2006,433(1-3):248-252.
[6]Li H.,Jiang T.,Zhang K."Efficient and robust feature extraction by
maximum margin criterion."IEEE Transactions on Neural Networks,2006,17(1):
157-165.
[7]Gunal S.,Gerek O.N.,Ece D.G.,Edizkan R."The search for optimal
feature set in power quality event classification."Expert Systems with
Applications,2009,36(7):10266-10273.
[8]Otsu N.A threshold selection method from gray-level histograms[J]
.IEEE Transactions on Systems,Man,and Cybernetics,1979,9(1):62-66.
[9]Wu J.,Wang Y.,Yu J.,Chen Y.,Pang Y.,Diao X.,Qiu Z."Identification
of implanted mesh after incisional hernia repair using an automated breast
volume scanner."Journal of Ultrasound in Medicine,2015,34(6):1071-1081.
[10]Kontos D.,Bakic P.R.,Carton A.K.,Troxel A.B.,Conant E.F.,Maidment
A.D."Parenchymal texture analysis in digital breast tomosynthesis for breast
cancer risk estimation:a preliminary study."Academic Radiology,2009,16(3):
283-298.。
Claims (9)
1. a kind of light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method, it is characterised in that comprise the following steps:
S1:Build a two dimension for ABUS coronal-planes(2-D)Foreground mask;
S2:Extract VOI;
S3:Pretreatment operation is carried out to VOI images;
S4:To giving one group of region to be sorted of VOI, can be to 40 features of each extracted region to be sorted;
S5:40 features to extracting are selected, and final selection makes light-type sticking patch and the error in classification of manadesma reach minimum
Feature as combinations of features;
It is described to build concretely comprising the following steps for 2-D foreground mask for coronal-plane in step S1:
S1.1:0.5 to 0.9 times of ABUS coronal-planes section of scanning total depth is located at by allC 1-C nTake out, to allC 1-C nFigure
The pixel of same position does average value processing as in, obtains a coronal-plane average imageC mean;
S1.2:Using Otsu algorithms to imageC meanThreshold process is carried out, bianry image is obtainedC binary;
S1.3:Using morphology opening operation pairC binaryImage is processed, and obtains coronal-plane 2-D foreground mask imagesC mask。
2. light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method according to claim 1, it is characterised in that
In step S2, the extraction VOI's concretely comprises the following steps:
S2.1:From ABUS coronal-plane mask imagesC maskThe upper left corner starts, and image is divided into the image block of 50 × 50 pixels;
S2.2:The coronal-plane position feature of all black picture blocks is set to 0, by all images intersected with foreground mask border
The coronal-plane position feature of block is set to 1, and the coronal-plane position feature of all full white image blocks is set into 2;
S2.3:One by one by all coronal-plane position features not for 0 image block is chosen for current interest region(ROI);Use
All ABUS cross sections related to ROI and sagittal plane image-region, the ROI of 2-D is expanded to the VOI of 3-D;One by one will be current
VOI sends into subsequent characteristics extraction module, until completing the traversal to all VOI.
3. light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method according to claim 2, it is characterised in that
It is described pretreatment operation is carried out to VOI images to concretely comprise the following steps in step S3:
S3.1:Use 3-D intelligence spot noise reduction anisotropy parameters(ISRAD)Algorithm carries out three-dimensional to the VOI for automatically extracting out
Filtering process;
S3.2:The single frames cross-sectional image for being pointed to VOI centers using Otsu algorithms does binary conversion treatment;
S3.3:White connected region using all areas in opening operation deletion bianry image less than 15 pixels;
S3.4:The minimum circumscribed bounding box of each white connected region is extracted, width, height and the summit for calculating bounding box are sat
Mark;
S3.5:Delete the width of bounding box<The white connected region of 15 pixels;
S3.6:Mark is numbered to all remaining white connected regions, as region to be sorted.
4. light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method according to claim 3, it is characterised in that
It is described feature extraction is carried out to each region to be sorted one by one to concretely comprise the following steps in step S4:
S4.1:The 2-D textural characteristics in region to be sorted are extracted, is carried out in two steps;The first step, calculates region cross section to be sorted list
12 2-D gray level co-occurrence matrixes of frame section(GLCM)Textural characteristics;Second step, calculates coronal-plane Slice Sequence in region to be sorted
12 2-D GLCM textural characteristics and 1 2-D fractal dimension(FD)Feature, is total up to 25;
S4.2:Extract the 3-D textural characteristics in region to be sorted;To the volume data in each region to be sorted, 12 3-D are calculated
GLCM textural characteristics and 1 3-D FD feature, are total up to 13;
S4.3:The local feature of scan depths residing for region to be sorted is extracted, it is deep based on being scanned residing for region to be sorted with one
Degree is the local feature f in the Y- directions of ABUSdepthTo characterize the occurrence probability of sticking patch;
S4.4:The environmental characteristic in region to be sorted and hernical sac position relationship is extracted, with one based on region to be sorted and hernical sac position
Put the location parameter f of relationadjacencyTo characterize the occurrence probability of sticking patch;
12 textural characteristics, including 2-D GLCM and 3-D GLCM, specially:Energy(f1), contrast(f2), it is related
(f3), variance(f4), homogeney(f5), average(f6), entropy(f7), auto-correlation(f8), otherness(f9), cluster shade(f10), collection
Group's protrusion(f11)And maximum probability(f12).
5. light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method according to claim 4, it is characterised in that
In step S5,40 features of described pair of extraction carry out concretely comprising the following steps for feature selecting:First, class spacing method is used(DBC)
The class spacing of each feature is calculated respectively;Secondly, being sorted from big to small by class spacing to 40 features, chooses class spacing larger
Preceding 25 features;Finally, the feature to tentatively select 25 uses sequential advancement search method(SFS)Selected, with
To making classification accuracy rate highest combinations of features.
6. light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method according to claim 4, it is characterised in that
In step S4.1,12 2-D gray level co-occurrence matrixes for calculating the section of region cross section to be sorted single frames(GLCM)Texture
Feature, concretely comprises the following steps:First, the single frames cross section noise-reduced image at VOI centers is extracted;Secondly, to each region to be sorted
The region extension of upper and lower 5 pixel is done in scan depths direction;Then, the cross section single frames for being partitioned into each region to be sorted is cut
Piece;Finally, 12 descriptors are calculated to single frames section, as the 2-D GLCM lines that the region cross section to be sorted single frames is cut into slices
Reason feature;
12 2-D GLCM textural characteristics for calculating coronal-plane Slice Sequence in region to be sorted, concretely comprise the following steps:For 2-D
GLCM, the differences in spatial location motion vector D of two pixels in two dimensional image(φ,d)To describe, d is two pixels
Between distance, φ is the angle of two pixels and reference axis;Set a distance d is given for one, on 4 independent directions:φ =
0th, 45,90,135, there may be 8 adjacent pixels altogether to occurring, then, the φ directions in two dimensional image are separated by d's
A pair of pixels pair, the probability for occurring with gray scale i and j respectively, i.e. p(i,j/φ,d), it is designated as pij;Will be by pijThe matrix of composition
Normalization obtains the gray level co-occurrence matrixes of image;Then, the calculating formula of 12 2-D GLCM textural characteristics is as follows:
Wherein, NgIt is grey level, p(i,j)It is gray scale pair(i,j)The normalized knot of the probability of appearance, i.e. gray level co-occurrence matrixes
Really;The average value and standard deviation of matrix row and column is,, and,;
1 2-D fractal dimension for calculating coronal-plane Slice Sequence in region to be sorted(FD)Feature, concretely comprises the following steps:For
2-D FD, two-dimensional fractal dimension is estimated by the power spectrum of image Fourier transformation;Use following Fast Fourier Transform (FFT)
(FFT)To carry out discrete Fourier transform to two dimensional image(DFT):
(13)
Wherein, I is that size is(M,N)Two dimensional image region, u and v is respectively the spatial frequency in x and y directions, u=0,
1 ... M-1, v=0,1 ... N-1;Power spectral density P passes through F(u,v)Estimate as follows:
(14)
To calculate 2-D FD, P is carried out averagely along the radial section direction across FFT frequency domains, frequency space is by equably decile
It is 24 directions, and 30 points of equably being sampled to the radial component in each direction;Calculate log(Pf)To log(f)Minimum
Two multiply fitting, whereinRadial frequency is represented, then FD is relevant to this double logarithmic curve in following form
Slope β:
(15)
Wherein, DTIt is topological dimension, for two dimensional image, DT = 2。
7. light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method according to claim 4, it is characterised in that
In step S4.2,12 3-D GLCM textural characteristics of the calculating are concretely comprised the following steps:For 3-D GLCM, in 3-D view
Two tissue points differences in spatial location motion vector D(φ,θ,d)To describe, d is the distance between two tissue points, and φ is
Azimuth between two tissue points, θ is the zenith angle between two tissue points;Set a distance d is given for one, in 13 independent directions
On, there may be 26 adjacent voxels altogether to occurring;Here, still selective extraction and 12 3-D of identical in 2-D GLCM
GLCM textural characteristics;
1 3-D FD feature of the calculating, concretely comprises the following steps:For 3-D FD, three-dimensional fractal dimension by volume images Fu 3-D
In the power spectrum of leaf transformation estimated;3-D DFT are carried out to whole 3-D view using following 3-D FFT:
(16)
Wherein,IIt is that size is(M,N,K)3-D view region,u, vWithwBe respectivelyx, yWithzThe spatial frequency in direction,
Power spectral densityPEstimate as follows:
(17)
To calculate 3-D FD, P is carried out averagely along the radial sector direction across 3-D FFT frequency domains;Frequency space is by equably
Decile is carried out in 24 azimuth directions and 12 zenith angular direction, and the radial component in each direction is equably sampled 30
Point, calculates log(Pf)To log(f)Least square fitting, whereinRepresent radial frequency,
Then 3-D FD are relevant to the slope β of this double logarithmic curve in following form:
(18)
Wherein, DTIt is topological dimension, for 3-D view, DT = 3。
8. light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method according to claim 4, it is characterised in that
In step S4.3, scan depths residing for region to be sorted are the local feature f in the Y- directions of ABUSdepthSet as follows:Edge
ABUS scan depths direction carries out interval division by probe interface to local feature to skin depth below:Respectively interval 0 ~ 1
Cm, 1 ~ 3 cm of interval and 3 ~ 6 cm of interval;According to the depth parameter in region to be sorted, to local feature fdepthSet:1)Interval 0
~ 1 cm, light-type sticking patch occurrence probability is medium on the upper side, i.e. fdepthIt is set to 3.; 2)1 ~ 3 cm of interval, light-type sticking patch goes out
Existing probability highest, i.e. fdepthIt is set to 4.;3)It is interval in 3 ~ 6 cm, with the increase of depth, place the possibility of sticking patch by
It is decrescence small, therefore from top to bottom, i.e. the f in this regiondepth2. and 1. it is set to.
9. light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method according to claim 4, it is characterised in that
In step S4.4, the environmental characteristic for extracting region to be sorted and hernical sac position relationship is divided to following two step to carry out:
The first step, detects location algorithm, to determine hernical sac position using the quick hernical sac based on ABUS data:First, detect
All black objects in ABUS coronal images;Secondly, after pseudo- black objects are filtered, maximum black target is calculated also just
It is size of the hernical sac in X-, Y- and Z- direction;3rd, to ensure that whole hernical sac is all contained among VOI volumes, by hernical sac size
All expand 40 pixels in three directions;4th, concentrated from ABUS volume datas and cut out the VOI volumes containing hernical sac;5th,
Spot noise reduction process is carried out to VOI using 3-D ISRAD algorithms;Finally, it is partitioned into hernical sac from every frame cross-sectional image of VOI
Profile, completes the detection to hernical sac in ABUS data and localization process;
Second step, after the completion of hernical sac detection, is adjusted to coronal-plane position feature:Pair with hernical sac coronal-plane view field weight
The coronal-plane position feature in folded or intersecting region to be sorted adds 2;Pair with the cm of hernical sac coronal-plane projected area distance 3 within treat
The coronal-plane position feature of specification area adds 1;The probability for sticking patch occur in the to be sorted region nearer apart from hernical sac position is just
It is bigger, therefore, with a location parameter f based on region to be sorted with hernical sac position relationshipadjacencyTo characterize the appearance of sticking patch
Probability;So, using the coronal-plane position feature value in region to be sorted as environmental characteristic fadjacencyValue;Area to be sorted
The f in domainadjacencyValue is bigger, and corresponding sticking patch occurrence probability is bigger.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710181583.6A CN106875409B (en) | 2017-03-24 | 2017-03-24 | A kind of light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710181583.6A CN106875409B (en) | 2017-03-24 | 2017-03-24 | A kind of light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106875409A true CN106875409A (en) | 2017-06-20 |
CN106875409B CN106875409B (en) | 2019-06-21 |
Family
ID=59172836
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710181583.6A Expired - Fee Related CN106875409B (en) | 2017-03-24 | 2017-03-24 | A kind of light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106875409B (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107610107A (en) * | 2017-09-01 | 2018-01-19 | 华中科技大学 | A kind of three-dimensional vascular plaque features of ultrasound pattern based on dimension describes method |
CN109300170A (en) * | 2018-10-18 | 2019-02-01 | 云南大学 | Portrait photo shadow transmission method |
CN110246125A (en) * | 2019-05-31 | 2019-09-17 | 天津大学 | Teat placement automatic testing method based on ABUS coronal image |
CN110991339A (en) * | 2019-12-02 | 2020-04-10 | 太原科技大学 | Three-dimensional puckery palate identification method adopting circular spectrum |
CN111210427A (en) * | 2020-01-17 | 2020-05-29 | 云南大学 | Time-change-based method for analyzing post-operation shrinkage of in-vivo light-weight patch |
CN111275617A (en) * | 2020-01-09 | 2020-06-12 | 云南大学 | Automatic splicing method and system for ABUS breast ultrasound panorama and storage medium |
CN111528836A (en) * | 2020-05-06 | 2020-08-14 | 北京工业大学 | Brain function network feature extraction method based on dynamic directional transfer function |
CN111626986A (en) * | 2020-04-29 | 2020-09-04 | 云南大学 | Three-dimensional ultrasonic abdominal wall hernia patch detection method and system based on deep learning |
CN116188488A (en) * | 2023-01-10 | 2023-05-30 | 广东省第二人民医院 | Gray gradient-based B-ultrasonic image focus region segmentation method and device |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104657984A (en) * | 2015-01-28 | 2015-05-27 | 复旦大学 | Automatic extraction method of three-dimensional breast full-volume image regions of interest |
-
2017
- 2017-03-24 CN CN201710181583.6A patent/CN106875409B/en not_active Expired - Fee Related
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104657984A (en) * | 2015-01-28 | 2015-05-27 | 复旦大学 | Automatic extraction method of three-dimensional breast full-volume image regions of interest |
Non-Patent Citations (3)
Title |
---|
JUN WU ET AL.: "A Fast Detection and Diagnosis Algorithm for Abdominal Incisional Hernia Lesions With Automated 3D Ultrasound Images", 《2013 6TH INTERNATIONAL CONFERENCE ON BIOMEDICAL ENGINEERING AND INFORMATICS (BMEI 2013)》 * |
JUN WU ET AL.: "Identification of Implanted Mesh After Incisional Hernia Repair Using an Automated Breast Volume Scanner", 《J ULTRASOUND MED》 * |
吴俊 等: "基于自动化三维超声冠状面图像的腹壁手术切口检测算法", 《仪器仪表学报》 * |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107610107A (en) * | 2017-09-01 | 2018-01-19 | 华中科技大学 | A kind of three-dimensional vascular plaque features of ultrasound pattern based on dimension describes method |
CN107610107B (en) * | 2017-09-01 | 2019-09-13 | 华中科技大学 | A kind of three-dimensional vascular plaque features of ultrasound pattern based on dimension describes method |
CN109300170A (en) * | 2018-10-18 | 2019-02-01 | 云南大学 | Portrait photo shadow transmission method |
CN110246125A (en) * | 2019-05-31 | 2019-09-17 | 天津大学 | Teat placement automatic testing method based on ABUS coronal image |
CN110991339A (en) * | 2019-12-02 | 2020-04-10 | 太原科技大学 | Three-dimensional puckery palate identification method adopting circular spectrum |
CN111275617A (en) * | 2020-01-09 | 2020-06-12 | 云南大学 | Automatic splicing method and system for ABUS breast ultrasound panorama and storage medium |
CN111275617B (en) * | 2020-01-09 | 2023-04-07 | 云南大学 | Automatic splicing method and system for ABUS breast ultrasound panorama and storage medium |
CN111210427A (en) * | 2020-01-17 | 2020-05-29 | 云南大学 | Time-change-based method for analyzing post-operation shrinkage of in-vivo light-weight patch |
CN111210427B (en) * | 2020-01-17 | 2023-03-10 | 云南大学 | Time-change-based method for analyzing post-operation shrinkage of in-vivo light-weight patch |
CN111626986A (en) * | 2020-04-29 | 2020-09-04 | 云南大学 | Three-dimensional ultrasonic abdominal wall hernia patch detection method and system based on deep learning |
CN111626986B (en) * | 2020-04-29 | 2022-06-21 | 云南大学 | Three-dimensional ultrasonic abdominal wall hernia patch detection method and system based on deep learning |
CN111528836A (en) * | 2020-05-06 | 2020-08-14 | 北京工业大学 | Brain function network feature extraction method based on dynamic directional transfer function |
CN111528836B (en) * | 2020-05-06 | 2023-04-28 | 北京工业大学 | Brain function network feature extraction method based on dynamic directional transfer function |
CN116188488A (en) * | 2023-01-10 | 2023-05-30 | 广东省第二人民医院 | Gray gradient-based B-ultrasonic image focus region segmentation method and device |
CN116188488B (en) * | 2023-01-10 | 2024-01-16 | 广东省第二人民医院(广东省卫生应急医院) | Gray gradient-based B-ultrasonic image focus region segmentation method and device |
Also Published As
Publication number | Publication date |
---|---|
CN106875409B (en) | 2019-06-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106875409B (en) | A kind of light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method | |
Tzikopoulos et al. | A fully automated scheme for mammographic segmentation and classification based on breast density and asymmetry | |
Tan et al. | Computer-aided detection of cancer in automated 3-D breast ultrasound | |
Lo et al. | Multi-dimensional tumor detection in automated whole breast ultrasound using topographic watershed | |
CN108364294A (en) | Abdominal CT images multiple organ dividing method based on super-pixel | |
US20050027188A1 (en) | Method and apparatus for automatically detecting breast lesions and tumors in images | |
CN101727537A (en) | Computer determining method of mammary gland CR image based on double visual angle information fusion | |
Biswas et al. | Mammogram classification using gray-level co-occurrence matrix for diagnosis of breast cancer | |
Ganvir et al. | Filtering method for pre-processing mammogram images for breast cancer detection | |
Jiang et al. | Segmentation of prostate ultrasound images: the state of the art and the future directions of segmentation algorithms | |
Suganya et al. | Feature extraction and classification of ultrasound liver images using haralick texture-primitive features: Application of SVM classifier | |
Pengiran Mohamad et al. | Transition of traditional method to deep learning based computer-aided system for breast cancer using Automated Breast Ultrasound System (ABUS) images: a review | |
Edwin et al. | Liver and tumour segmentation from abdominal CT images using adaptive threshold method | |
Pashoutan et al. | Automatic breast tumor classification using a level set method and feature extraction in mammography | |
Wang et al. | Automatic detection of regions of interest in breast ultrasound images based on local phase information | |
Abdellatif et al. | K2. Automatic pectoral muscle boundary detection in mammograms using eigenvectors segmentation | |
Rodrigues et al. | A new methodology based on q-entropy for breast lesion classification in 3-D ultrasound images | |
da Cruz | Automatic analysis of mammography images | |
Tzikopoulos et al. | A fully automated complete segmentation scheme for mammograms | |
Timp | Analysis of Temporal Mammogram Pairs to Detect and Characterise Mass Lesions. | |
da Cruz | Automatic analysis of mammography images: enhancement and segmentation techniques | |
Kekre et al. | Texture based segmentation using statistical properties for mammographic images | |
Ciurte et al. | Texture analysis within contrast enhanced abdominal ct images | |
US11954853B2 (en) | Systems and methods for fast mammography data handling | |
Zhisheng et al. | A Novel Coarse-to-Fine Segmentation Method for Pancreatic Cancer |
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 | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190621 Termination date: 20200324 |
|
CF01 | Termination of patent right due to non-payment of annual fee |