CN106651846B - Segmentation method of retinal blood vessel image - Google Patents
Segmentation method of retinal blood vessel image Download PDFInfo
- Publication number
- CN106651846B CN106651846B CN201611185885.2A CN201611185885A CN106651846B CN 106651846 B CN106651846 B CN 106651846B CN 201611185885 A CN201611185885 A CN 201611185885A CN 106651846 B CN106651846 B CN 106651846B
- Authority
- CN
- China
- Prior art keywords
- image
- matched filtering
- pixel
- blood vessel
- thick
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 79
- 230000011218 segmentation Effects 0.000 title claims abstract description 40
- 210000001210 retinal vessel Anatomy 0.000 title claims abstract description 22
- 210000004204 blood vessel Anatomy 0.000 claims abstract description 87
- 238000001914 filtration Methods 0.000 claims abstract description 65
- 230000004044 response Effects 0.000 claims abstract description 35
- 238000012545 processing Methods 0.000 claims abstract description 12
- 230000002207 retinal effect Effects 0.000 claims description 29
- 230000002792 vascular Effects 0.000 claims description 27
- 230000008569 process Effects 0.000 claims description 11
- 238000006243 chemical reaction Methods 0.000 claims description 7
- 239000000284 extract Substances 0.000 claims description 5
- 238000003709 image segmentation Methods 0.000 claims description 2
- 230000002787 reinforcement Effects 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 claims 1
- 210000001525 retina Anatomy 0.000 abstract description 8
- 230000003044 adaptive effect Effects 0.000 abstract description 2
- 239000008280 blood Substances 0.000 description 9
- 210000004369 blood Anatomy 0.000 description 9
- 230000004927 fusion Effects 0.000 description 7
- 238000001514 detection method Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 230000003902 lesion Effects 0.000 description 5
- 235000013399 edible fruits Nutrition 0.000 description 4
- 230000004256 retinal image Effects 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000002902 bimodal effect Effects 0.000 description 2
- 230000015624 blood vessel development Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 201000010099 disease Diseases 0.000 description 2
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 210000005166 vasculature Anatomy 0.000 description 2
- 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
- 241001632422 Radiola linoides Species 0.000 description 1
- 208000034189 Sclerosis Diseases 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000036772 blood pressure Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 206010012601 diabetes mellitus Diseases 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 206010020718 hyperplasia Diseases 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000002401 inhibitory effect Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 210000003733 optic disk Anatomy 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000003014 reinforcing effect Effects 0.000 description 1
- 238000006467 substitution reaction 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
-
- 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
- G06T2207/10024—Color 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
- G06T2207/20024—Filtering details
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30041—Eye; Retina; Ophthalmic
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The embodiment of the invention provides a segmentation method of a retinal blood vessel image. The method mainly comprises the following steps: performing double-scale matched filtering processing on the retinal vessel image to obtain a fine-scale matched filtering response image and a coarse-scale matched filtering response image; segmenting line support areas from the fine-scale matched filtering response image, and performing binarization processing on each line support area by using a local adaptive threshold method to segment a fine blood vessel section; and (4) segmenting the coarse-scale matched filtering image by using a fixed proportion threshold algorithm to obtain a coarse vessel section. And the segmentation result of the thin blood vessel section and the segmentation result of the thick blood vessel section are fused to obtain a complete retina blood vessel segmentation result, and the accuracy of the segmentation result is high.
Description
Technical field
The present invention relates to digital image processing techniques field more particularly to a kind of dividing methods of retinal vascular images.
Background technique
Retinal vessel is the blood vessel that noninvasive can uniquely observe directly in Whole Body blood vessel, its shape, caliber,
Whether scale, Branch Angle change, and whether have hyperplasia, exudation, can reflect the lesion of system vascular.Diabetes, height
The diseases such as blood pressure, cerebrovascular sclerosis can lead to retinal vessel and occur certain variation, thus the detection of retinal vessel and
Analysis clinically has important directive significance to the diagnosing and treating of these diseases.However, due to retinal vascular images
Intensity profile it is uneven, target blood is low with the contrast of image background, along with the pollution of picture noise, makes retinal blood
The segmentation automatically of pipe is extremely difficult.
Currently, a kind of dividing method of retinal vascular images in the prior art are as follows: retinal vessel method for tracing.
Tolias (1998) etc. applies fuzzy C-mean algorithm method, establishes seed point from retina disk place, utilizes the one of blood vessel section
The tracking that fuzzy model carries out retinal vessel is tieed up, is mentioned by determining whether all seed points belong to the final segmentation of blood vessel progress
It takes.This method can be than more fully describing retinal vessel network structure.
The shortcomings that above-mentioned retinal vessel method for tracing are as follows: operand is big, and branch point and comparison for blood vessel
It is inaccurate to spend lower blood vessel segmentation.This method is more sensitive to the selection of seed point, and algorithm is often in vessel branch point
Place terminates, and loses a large amount of small retinal blood vessels.
The dividing method of another kind retinal vascular images in the prior art are as follows: classifier recognition methods.This method master
If establishing feature space according to the different characteristics of blood vessel, selection is suitable by carrying out pretreatment operation to retinal vessel
Classifier is trained sample, then brings into test set and determines last result.Staal (2004) uses ridge detection
Method has carried out supervision class using the feature vector of vessel centerline neighborhood and has identified.Soares (2006) extracts retina first
Then the multiple dimensioned 2D Gabor wavelet feature of blood-vessel image identifies retinal vessel using Bayes classifier.
The shortcomings that above-mentioned classifier recognition methods are as follows: this classification method is more sensitive to noise spot, and segmentation result
Existing misclassification situation is serious.
Summary of the invention
The embodiment provides a kind of dividing methods of retinal vascular images, to realize effectively from retina
Thick vessel segment and thin vessel segment are partitioned into blood-vessel image.
To achieve the goals above, this invention takes following technical solutions.
A kind of dividing method of retinal vascular images, comprising:
Double scale matched filterings processing are carried out to retinal vascular images, obtain thin scale matched filtering response image and thick
Scale matched filtering response image；
Divide outlet support area from the thin scale matched filtering response image, uses local auto-adaptive threshold method
Binary conversion treatment is carried out to each line support area, is partitioned into thin vessel segment；
The thick scale matched filtering image is split using fixed proportion thresholding algorithm, obtains thick vessel segment.
Further, described that double scale matched filtering processing are carried out to retinal vascular images, obtain thin scale matching
Filter response image and thick scale matched filtering response image, comprising:
Extract the green channel in tri- channels RGB of color retinal vascular images；
The cross section grey scale curve that retinal vessel is simulated with Gaussian function, obtains following matched filter:
In formula, K (x, y) is referred to as kernel function, and σ is irrelevance of the Gaussian function along x-axis coordinate center, and L is Gaussian function
The lightning channel length being truncated along y-axis, x in formula, y need to meet | x |≤3 σ, | y |≤L/2；
With 15 ° for interval, 12 directions in angular interval [0 °, 180 °] are chosen, 12 matched filters are created；
Green channel in the color retinal vascular images is made into convolution meter with 12 matched filters respectively
It calculates, obtains matched filtering response image, the matched filtering response image is normalized and be quantified as 256 grades of grayscale image, when
When the irrelevance σ is less than the threshold value of setting, using obtained grayscale image as thin scale matched filtering response image；When described inclined
When from degree σ not less than the threshold value set, using obtained grayscale image as thick scale matched filtering response image.
It is further, described to divide outlet support area from the thin scale matched filtering response image, comprising:
The gradient magnitude and gradient direction for calculating each pixel in thin scale matched filtering image, by all pixels point
It is ranked up according to its gradient magnitude size, choosing has the pixel of highest gradient magnitude as seed point, by gradient magnitude
Pixel less than the Grads threshold of setting excludes outside the building process of online support area；
Several line support areas are generated using algorithm of region growing based on the seed point, each line support area includes
One seed point, and be a pixel set with seed point with similar gradient direction, each pixel includes two shapes
State: using and is not used.
Further, described that several line support areas are generated using algorithm of region growing based on the seed point, it wraps
It includes:
Select a not used pixel as seed point from the sorted lists of pixel, by the ladder of the seed point
Spend initial angle θ of the direction as the line support area where the seed point to be generatedregion, by the neighbour of the seed point
Not used and its gradient direction is with regional perspective θ in domainregionBetween pixel of the error between τ be added to the line
In support area, the angle for calculating the line support area is updated according to updated pixel, wherein τ is angle threshold；
Above-mentioned treatment process is repeated, until there is no qualified pixel to be added in the seed neighborhood of a point
In the line support area, the corresponding minimum circumscribed rectangle in the line support area is extended.
Further, the use local auto-adaptive threshold method carries out at binaryzation each line support area
Reason, is partitioned into thin vessel segment, comprising:
After thin scale matched filtering image segmentation is gone out multiple line support areas, local auto-adaptive threshold method application is used
Otsu algorithm carries out binary conversion treatment to each line support area, is partitioned into foreground and background, is partitioned into single thin blood vessel
Section；
The Otsu method searches for optimal threshold value and makes the variance between prospect and background maximum, if t is prospect and back
The segmentation threshold of scape then calculates the probability w of foreground pixel0tWith average gray u0t, the probability w of background pixel1tIt is with average gray
u1t, the variance between foreground and background indicates are as follows:
gt=w0t·(u0t-ut)2+w1t·(u1t-ut)2,
Wherein utIndicate image overall average gray scale, the value range of t is 0-255, as variance gtWhen maximum, foreground and background
Difference is maximum, then corresponding gray scale t is optimal threshold.
Further, the application fixed proportion thresholding algorithm is split the thick scale matched filtering image,
Obtain thick vessel segment, comprising:
The thick scale matched filtering image is split using fixed proportion thresholding algorithm, obtains thick blood-vessel image,
The threshold value of the fixed proportion thresholding algorithm is calculated by following formula:
Wherein r is input parameter, indicates expected blood vessel ratio, and Num is that the frequency calculates function, and Total indicates that pixel is total
Number；
In the application fixed proportion thresholding algorithm, first the pixel in the thick scale matched filtering image is dropped
Sequence sequence, searches for optimal threshold value Tr, carries out binaryzation to the thick scale matched filtering image according to the optimal threshold Tr
Processing, is partitioned into foreground and background, is partitioned into single thick vessel segment.
Further, the method further include:
The segmentation result of the thick vessel segment and the segmentation result of the thin vessel segment are melted using logic or operation
It closes, obtains the segmentation result of complete thick vessel segment and thin vessel segment.
As can be seen from the technical scheme provided by the above-mentioned embodiment of the present invention, the method for the embodiment of the present invention passes through ALT
Method can effectively be partitioned into thin blood vessel from retinal vascular images, can be effectively from retinal blood by FRT method
It is partitioned into complete thick blood vessel in pipe image, merges ALT method and the available complete retinal vessel segmentation knot of FRT method
Fruit, segmentation result accuracy rate are high.
The additional aspect of the present invention and advantage will be set forth in part in the description, these will become from the following description
Obviously, or practice through the invention is recognized.
Detailed description of the invention
In order to illustrate the technical solution of the embodiments of the present invention more clearly, required use in being described below to embodiment
Attached drawing be briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, for this
For the those of ordinary skill of field, without any creative labor, it can also be obtained according to these attached drawings other
Attached drawing.
Fig. 1 is a kind of retina of double scale non-linear thresholds based on blood vessel support area provided in an embodiment of the present invention
The realization principle schematic diagram of the dividing method of blood-vessel image；
Fig. 2 is a kind of retina of double scale non-linear thresholds based on blood vessel support area provided in an embodiment of the present invention
The process flow diagram of the dividing method of blood-vessel image；
Fig. 3 is a kind of rectangular extension figure provided in an embodiment of the present invention, wherein figure (a) indicates what region growing algorithm obtained
Rectangle, figure (b) indicate the extension of rectangle in abscissa and ordinate both direction, and figure (c) indicates the rectangle after extension；
Fig. 4 is a kind of line support area grey level histogram provided in an embodiment of the present invention；
Fig. 5 is a kind of thickness blood vessel fusion figure provided in an embodiment of the present invention, and figure (a) indicates the inspection of local auto-adaptive threshold value
Result example is surveyed, figure (b) indicates that the segmentation result of fixed proportion thresholding algorithm, figure (c) indicate final fusion results.
Specific embodiment
Embodiments of the present invention are described below in detail, the example of the embodiment is shown in the accompanying drawings, wherein from beginning
Same or similar element or element with the same or similar functions are indicated to same or similar label eventually.Below by ginseng
The embodiment for examining attached drawing description is exemplary, and for explaining only the invention, and is not construed as limiting the claims.
Those skilled in the art of the present technique are appreciated that unless expressly stated, singular " one " used herein, " one
It is a ", " described " and "the" may also comprise plural form.It is to be further understood that being arranged used in specification of the invention
Diction " comprising " refer to that there are the feature, integer, step, operation, element and/or component, but it is not excluded that in the presence of or addition
Other one or more features, integer, step, operation, element, component and/or their group.It should be understood that when we claim member
Part is " connected " or when " coupled " to another element, it can be directly connected or coupled to other elements, or there may also be
Intermediary element.In addition, " connection " used herein or " coupling " may include being wirelessly connected or coupling.Wording used herein
"and/or" includes one or more associated any cells for listing item and all combinations.
Those skilled in the art of the present technique are appreciated that unless otherwise defined, all terms used herein (including technology
Term and scientific term) there is meaning identical with the general understanding of those of ordinary skill in fields of the present invention.Also answer
It should be appreciated that those terms such as defined in the general dictionary should be understood that have in the context of the prior art
The consistent meaning of meaning, and unless defined as here, it will not be explained in an idealized or overly formal meaning.
In order to facilitate understanding of embodiments of the present invention, it is done by taking several specific embodiments as an example below in conjunction with attached drawing further
Explanation, and each embodiment does not constitute the restriction to the embodiment of the present invention.
The embodiment of the present invention proposes a kind of retinal vessel of double scale non-linear thresholds based on blood vessel support area
The dividing method of image, this method use the Gaussian filter of two different scales to carry out eye ground blood-vessel image first
Pretreatment operation obtains the matched filtering response image of two different scales.Then respectively to the matching of the two different scales
Filter response image carries out retinal vessel extraction.Finally the blood vessel extracted is merged, obtains final result.Specifically
Steps are as follows:
The embodiment of the present invention proposes a kind of retinal vessel of double scale non-linear thresholds based on blood vessel support area
The realization principle schematic diagram of the dividing method of image is as shown in Figure 1, specifically process flow is as shown in Fig. 2, include following processing
Step:
Step S210, to retinal vascular images carry out double scale matched filterings (double-scale fi1tering,
DSF it) handles, obtains thin scale matched filtering response (fine matched filtering response, FMFR) image and thick
Matched filtering responds (coarse matched fi1tering response, CMFR) image.
The Gauss matched filtering device of different scale is different for the reinforcing effect of thin blood vessel and thick blood vessel.More specifically
, small scaling filter is conducive to protrude thin blood vessel, extracts the skeleton of thick blood vessel；And thick scaling filter is conducive to enhance thick blood
Pipe, is blurred the thin blood vessel of tip.Therefore, we devise two class gaussian kernel function of thin scale and thick scale, they are distinguished
The green channel for acting on retinal vascular images obtains thin scale matched filtering response image and thick scale matched filtering response
Image.
In three channels RGB (Red, Green, Blue) of color retinal vascular images, green channel component image
Blood vessel and background contrasts highest, be more conducive to blood vessel segmentation, therefore the present invention extracts color retinal vascular figure first
The green channel of picture.
In retinal vascular images, blood vessel center pixel brightness is smaller, and the pixel brightness on both sides is larger, retina
The cross section gray level skeleton of blood vessel can be approximate with Gaussian.Therefore Gauss matched filtering method is commonly used to promote image pair
Degree of ratio.It is assumed that retinal vessel is the wide straightway of segmentation, the length is L, width is 2 σ, we are simulated with Gaussian function
The cross section grey scale curve of retinal vessel, to obtain following matched filter:
In formula, K (x, y) is referred to as kernel function, and σ is irrelevance of the Gaussian function along x-axis coordinate center, and L is Gaussian function
The lightning channel length being truncated along y-axis, in order to enable matching is more accurate, x in formula, y need to meet | x |≤3 σ, | y |≤L/2.Root
It factually tests as a result, we set L=7.
Because vessel directions are arbitrary, we are 12 sides in interval consideration angular interval [0 °, 180 °] with 15 °
To creating 12 matched filters.
Retinal vascular images do convolution, each of matched filtering response image pixel with this 12 Gaussian kernels respectively
Value be equal to maximum convolution value.For the ease of subsequent processing, matched filtering response image is normalized and is quantified as 256 grades
Grayscale image.
It also include the thin blood of tip we can observe that both having included the thick blood vessel of optic disk attachment in retinal vascular images
Pipe.In application matched filtering, if selecting the σ of smaller numberical range, this comparison small range is that blood mentioned above is inclined
Gaussian function is along the irrelevance at x-axis coordinate center in other words for width from degree pipe, and about 1.3~1.6 pixels, then filter knot
Thin blood vessel in fruit image is more readily available reinforcement, and thick blood vessel is corroded；On the contrary, if selecting the σ of bigger numberical range,
This bigger numberical range is 2.0~2.4 pixels, then thick blood vessel is strengthened, and thin blood vessel is blurred.
Therefore, the embodiment of the present invention proposes double scale matched filtering methods.Thin scale matched filter selection is lesser
σ enhances thin blood vessel, while inhibiting noise and smooth background area.The response results that thin scale matched filter generates are through excess
Change, obtain FMFR image, the image by be auto-thresholding algorithm input.On the contrary, thick scale matched filter selection
Biggish σ enhances thick vasculature part, obtains CMFR image.Because the marginal portion of thick blood vessel is in thin scale matched filtering image
In be easy to be corroded, but thick blood vessel is easier the full segmentation from thick scale matched filtering image and comes out, so, CMFR figure
As segmentation result fusion will be used for.
Step S220, divide outlet support area (vesselsupport from thin scale matched filtering response image
Region, VSR), it is calculated using local auto-adaptive threshold value (adaptive local thresholding, ALT) method application Otsu
Method carries out binary conversion treatment to each VSR, is partitioned into single thin vessel segment.
After matched filtering, the contrast of FMFR image is enhanced, and especially stain and lesion region obtains
Inhibit.But the intensity profile of FMFR image medium vessels or more dispersed, gray value and the background gray levels of part blood vessel are deposited
In greater overlap.Theoretically, we can not find a global threshold linear partition blood vessel and background.VSR refers to comprising one
The rectangular area of vessel segment can be detected automatically by algorithm.In a part VSR, histogram has apparent bimodulus
State property matter, we can be using automatic threshold algorithm (such as Otsu) segmentation blood vessel and background.The process detects automatically first
Then each region VSR is divided using 0tsu algorithm in all regions VSR in FMFR, while the picture in all non-regions VSR
Element is set as background, finally obtains thin blood vessel segmentation figure (fine vessel segmentation, FVS).
S2-1: line support area detection
Since the thin blood vessel of retina is more sensitive to small scale parameter σ, so the present invention is calculated first in scale σ=1.3
Under the conditions of matched filtering image, then carry out the following processing on this basis.
S2-1-1: line support area generates
The gradient magnitude and gradient direction of each pixel in thin scale matched filtering image are calculated first, then by institute
There is pixel to be ranked up according to its gradient magnitude size.Stronger marginal point or region generally all have relatively high gradient width
Value usually has highest gradient magnitude in the pixel of retinal blood tube edges, therefore is chosen first with highest gradient magnitude
Pixel as seed point.In calculating process, the pixel that gradient magnitude is less than q (present invention selects 0.4) will be rejected
Participate in the building process of line support area.
The present invention generates several line support areas, each line support area namely one and kind using algorithm of region growing
Son point has the pixel set in similar gradient direction.Each pixel includes two states, even if used and unused.Initial shape
All pixels point is all set to unused by state.
Algorithm of region growing selects a not used pixel as seed point, the neighbour of the pixel from sorted lists first
Not used and its gradient direction is with regional perspective θ in domainregionBetween pixel of the error between τ will be added into the area
In domain.For the range of test τ in text probably between 18 ° to 24 °, general default takes 22.
The initial angle θ in regionregionIt is exactly the gradient direction of seed point, one new pixel of each addition to the region,
The angle in region is just updated.The angle in region is just updated to:
θjIndicate the vertical direction of pixel j gradient.
I (x, y) indicates the gray value at pixel (x, y) point, gx(x, y), gy(x, y) respectively indicate pixel (x, y) in x and
The gradient value in the direction y.
It so successively carries out, until no any pixel can be added in rectangle.
The line support area being previously obtained is indicated with a minimum circumscribed rectangle.So as to obtain some of rectangle
Essential information, Fig. 3 are a kind of rectangular extension figure provided in an embodiment of the present invention, coordinate and rectangle including rectangular centre point
Length and width and principal direction, wherein scheming (a) indicates that the rectangle that region growing algorithm obtains, figure (b) indicate abscissa and indulge
The extension of rectangle in coordinate both direction, figure (c) indicate the rectangle after extension.
The extension of rectangle is divided into two steps here, step 1: first horizontal and vertical to rectangle all carry out equivalent extension, extension
Amplitude is chosen for the half of rectangle width width).After having extended a rectangle every time in this way, by the shape of the pixel in rectangle
State is both configured to Used, these next points for being arranged to Used would not be selected as seed point.Step 2: again to expansion before
On the rectangular foundation of exhibition, equivalent extension is all carried out to horizontal and vertical, extended amplitude is as definite value size before.It is extending
Threshold process is carried out on rectangular foundation afterwards, can thus solve the problems, such as not intersect between two rectangles, because second
Those of increase pixel can be selected as seed point after secondary extension.
1. algorithm of region growing of algorithm
S2-1-2: rectangle growth
The line support area that algorithm of region growing obtains can cover most of lightning channel region well, but it exists
Both sides is insufficient: (1) gradient value of lightning channel two sides is bigger, thus the seed point that region increases is generally lightning channel
The pixel at two edges causes to will appear two rectangles in the same cross-section of lightning channel；(2) next region increases
Seed point, which is not necessarily, to be selected from generated rectangular area, thus newly-generated region keeps up with a region and not necessarily has
Overlapping region.Namely may be discontinuous along lightning channel direction rectangle, to all cannot include by all foreground areas
Enter.
Therefore, the invention proposes rectangular extension algorithms, i.e., two-way extension is carried out on original rectangular foundation, extend width
Degree is chosen for the half of rectangle width.Wherein it is discontinuous to can solve rectangular area for Longitudinal Extension (along the direction of blood vessel development)
The problem of, it is extending transversely (vertical with blood vessel development direction) to can achieve the effect for merging two rectangles.
Fig. 3 gives the schematic diagram of rectangular extension, and wherein O point is rectangular centre point, and theta is the principal direction of rectangle,
Length and Width is respectively the length and width of rectangle, and P, Q are respectively the central point on rectangle is wide, can be according to x1 and x2 and y1
It is divided into 9 kinds of situations with the size of y2.
S2-2: line support area Threshold segmentation
Using above-mentioned VSR detection algorithm, a width MFR image can be partitioned into multiple VSR.
Step S230, then ALT method application Otsu algorithm carries out binary conversion treatment to each VSR, is partitioned into prospect
And background, it is partitioned into single thin vessel segment.Intuitively, each region VSR is contrast obviously image block, blood
Pipeline section and background have significant difference in gray space.From statistics angle analysis, the intensity Distribution value in the region VSR has bimodal
Property, i.e., respective intensity interval is distributed in background pixel and blood vessel set of pixels.Fig. 4 illustrates two VSR's randomly selected
Grey level histogram.Our more experimental results all show that the intensity distribution of VSR has bimodal nature.
Otsu is a kind of automatic threshold method of classics, it has preferable segmentation effect to the image with bi-modal distribution
Fruit.The principle of Otsu method is that the optimal threshold value of search makes the variance between prospect and background maximum.Assuming that t is prospect and back
The segmentation threshold of scape can then calculate the probability w of foreground pixel0tWith average gray u0t, the probability w of background pixel1tWith average ash
Degree is u1t.Variance between foreground and background may be expressed as:
gt=w0t·(u0t-ut)2+w1t·(u1t-ut)2,
Wherein utIndicate that image overall average gray scale, the value range of t are 0-255.As variance gtWhen maximum, foreground and background
Difference is maximum, then corresponding gray scale t is optimal threshold.
Step S240, divide CMFR image using fixed proportion thresholding algorithm, obtain thick blood vessel segmentation figure, merge thickness blood
Pipe dividing method obtains the segmentation result of complete thin blood vessel and thick blood vessel.
Fusion includes two key steps: firstly, dividing CMFR image using fixed proportion thresholding algorithm, obtaining thick blood vessel
Segmentation figure (coarse vessel segmentation, CVS).Then, FVS and CVS is merged by logic or operation,
So that fusion results had both included thin blood vessel, also comprising complete thick blood vessel.
ALT can detecte to obtain thin blood-vessel image, but the thick blood vessel of ALT segmentation often only includes its skeleton, and omit
The pixel of its peripheral part.Fig. 5 (a) illustrates the testing result example of ALT.In order to improve the detection performance of ALT, the present invention
It is proposed the fusion method of thin blood vessel and thick blood vessel.This method applies fixed proportion thresholding algorithm (Fixed-ratio first
Thresholding, FRT) thick scale matched filtering image is split, thick blood-vessel image is obtained, thin vessel graph is then merged
Picture and thick blood-vessel image obtain final thickness blood vessel fusion results.
Fixed proportion thresholding algorithm is one kind simply based on the binarization method of priori.Intuitively, retinal images have
Have apparent structural, i.e., it is made of linear blood vessel and flat background, and the ratio of vasculature part is often relatively low.
From statistics angle analysis, the average value and variance of DRIVE medium vessels pixel ratio are respectively 8.43% and 1.38%, in STARES
The average value and variance of blood vessel pixel ratio are respectively 7.6% and 3.15%.2 data sets DRIVE and STARES are only with flat
Mean value removes variance, the average value 12.7% of DRIVE, the average value 10.4% of STARES as evaluation index.
It should be mentioned that having the retinal images of some lesions in STARES, so its blood vessel variance of proportion compares
Greatly.In thick scale matched filtering image, the response of thick blood vessel pixel is bigger than the response of thin blood vessel and background pixel.Cause
This, the threshold value of FRT method can be calculated by following formula:
Wherein r is input parameter, indicates expected blood vessel ratio.Num is that the frequency calculates function, and Total indicates that pixel is total
Number.When realizing the algorithm, descending sort first is carried out to pixel in thick scale matched filtering image using bucket sort algorithm, then
Optimal threshold value Tr is searched for, binaryzation is finally carried out to thick scale matched filtering image according to Tr.The segmentation of Fig. 5 (b) displaying FRT
As a result, it can be observed that the result is more completely partitioned into thick blood vessel, although it is lost the details of thin blood vessel.
To sum up, ALT is good at dividing thin blood vessel, and FRT can then be partitioned into complete thick blood vessel.Therefore, two methods are merged
Result be expected fairly perfect segmentation performance.Simply, present invention application logic or operation for ALT and FRT knot
Fruit is merged.It is 255 that the gray value of corresponding position pixel, which has one, i.e. in two images, then corresponding to position in result images
The gray value for setting pixel is 255, only when the gray value of corresponding position pixel in two images is 0, result images
The gray value of middle corresponding position pixel is just 0.
In addition, in order to eliminate the interference of ambient noise and parts of lesions tissue, area of the removal area less than 10 pixels
Domain.Fig. 5 (c) illustrates final fusion results.
In conclusion the method for the embodiment of the present invention can effectively be divided from retinal vascular images by ALT method
Thin blood vessel is cut out, complete thick blood vessel can be effectively partitioned into from retinal vascular images by FRT method, merge ALT
Method and the available complete retinal vessel segmentation result of FRT method, segmentation result accuracy rate are high.
Most of method all carries out vessel extraction just for normal, the preferable retinal images of imaging both at home and abroad at present,
And for the retinal images of low contrast or generation lesion, since blood vessel and background area pixels gray value size more connect
Closely, existing conventional method largely can not correctly split blood vessel with background.And the present invention utilizes size scale
Blood vessel is divided into thickness blood vessel by Gauss matched filtering, carries out enhancing processing respectively, and effect is obvious.For thin blood vessel, picture is utilized
The gradient magnitude of vegetarian refreshments and direction, are increased and the method for rectangular extension based on region, have preferable local auto-adaptive, can be with
Rapidly determine foreground area.And for thick blood vessel, using global fixed threshold, it can preferably retain trunk portion.
Those of ordinary skill in the art will appreciate that: attached drawing is the schematic diagram of one embodiment, module in attached drawing or
Process is not necessarily implemented necessary to the present invention.
As seen through the above description of the embodiments, those skilled in the art can be understood that the present invention can
It realizes by means of software and necessary general hardware platform.Based on this understanding, technical solution of the present invention essence
On in other words the part that contributes to existing technology can be embodied in the form of software products, the computer software product
It can store in storage medium, such as ROM/RAM, magnetic disk, CD, including some instructions are used so that a computer equipment
(can be personal computer, server or the network equipment etc.) executes the certain of each embodiment or embodiment of the invention
Method described in part.
All the embodiments in this specification are described in a progressive manner, same and similar portion between each embodiment
Dividing may refer to each other, and each embodiment focuses on the differences from other embodiments.Especially for device or
For system embodiment, since it is substantially similar to the method embodiment, so describing fairly simple, related place is referring to method
The part of embodiment illustrates.Apparatus and system embodiment described above is only schematical, wherein the conduct
The unit of separate part description may or may not be physically separated, component shown as a unit can be or
Person may not be physical unit, it can and it is in one place, or may be distributed over multiple network units.It can root
According to actual need that some or all of the modules therein is selected to achieve the purpose of the solution of this embodiment.Ordinary skill
Personnel can understand and implement without creative efforts.
The foregoing is only a preferred embodiment of the present invention, but scope of protection of the present invention is not limited thereto,
In the technical scope disclosed by the present invention, any changes or substitutions that can be easily thought of by anyone skilled in the art,
It should be covered by the protection scope of the present invention.Therefore, protection scope of the present invention should be with scope of protection of the claims
Subject to.
Claims (4)
1. a kind of dividing method of retinal vascular images characterized by comprising
Double scale matched filtering processing are carried out to retinal vascular images, obtain thin scale matched filtering response image and thick scale
Matched filtering response image, comprising: in application matched filtering, select the σ of smaller numberical range, this smaller range
It is 1.3~1.6 pixels, then the thin blood vessel in filter result image is more readily available reinforcement, and thick blood vessel is corroded；On the contrary,
The σ of bigger numberical range is selected, this bigger numberical range is 2.0~2.4 pixels, then thick blood vessel is strengthened, carefully
Blood vessel is blurred, wherein σ is width irrelevance of the Gaussian function along x-axis coordinate center in other words of blood vessel irrelevance；
Divide outlet support area from the thin scale matched filtering response image, using local auto-adaptive threshold method to every
One line support area carries out binary conversion treatment, is partitioned into thin vessel segment, comprising:
The gradient magnitude and gradient direction for calculating each pixel in thin scale matched filtering image, by all pixels point according to
Its gradient magnitude size is ranked up, and choosing has the pixel of highest gradient magnitude as seed point, and gradient magnitude is less than
The pixel of the Grads threshold of setting excludes outside the building process of online support area；Region growing is utilized based on the seed point
Algorithm generates several line support areas, and each line support area includes a seed point, and has for one with seed point
The pixel set in similar gradient direction, each pixel include two states: using and be not used；
After thin scale matched filtering image segmentation is gone out multiple line support areas, local auto-adaptive threshold method application Otsu is used
Algorithm carries out binary conversion treatment to each line support area, is partitioned into foreground and background, is partitioned into single thin vessel segment；
The Otsu method searches for optimal threshold value and makes the variance between prospect and background maximum, if t is foreground and background
Segmentation threshold then calculates the probability w of foreground pixel0tWith average gray u0t, the probability w of background pixel1tIt is u with average gray1t,
Variance between foreground and background indicates are as follows:
gt=wot·(u0t-ut)2+w1t·(u1t-ut)2
Wherein utIndicate image overall average gray scale, the value range of t is 0-255, as variance gtWhen maximum, foreground and background difference
Maximum, then corresponding gray scale t is optimal threshold；
The thick scale matched filtering image is split using fixed proportion thresholding algorithm, obtains thick vessel segment, comprising: answer
The thick scale matched filtering image is split with fixed proportion thresholding algorithm, obtains thick blood-vessel image, the fixed ratio
The threshold value of example thresholding algorithm is calculated by following formula:
Wherein r is input parameter, indicates expected blood vessel ratio, and Num is that the frequency calculates function, and Total indicates sum of all pixels；
In the application fixed proportion thresholding algorithm, descending row first is carried out to the pixel in the thick scale matched filtering image
Sequence searches for optimal threshold value Tr, carries out binary conversion treatment to the thick scale matched filtering image according to the optimal threshold Tr,
It is partitioned into foreground and background, is partitioned into single thick vessel segment.
2. the method according to claim 1, wherein described carry out double scale matchings to retinal vascular images
Filtering processing, obtains thin scale matched filtering response image and thick scale matched filtering response image, comprising:
Extract the green channel in tri- channels RGB of color retinal vascular images；
The cross section grey scale curve that retinal vessel is simulated with Gaussian function, obtains following matched filter:
In formula, K (x, y) is referred to as kernel function, and σ is irrelevance of the Gaussian function along x-axis coordinate center, and L is Gaussian function along y-axis
The lightning channel length being truncated, x in formula, y need to meet | x |≤3 σ, | y |≤L/2；
With 15 ° for interval, 12 directions in angular interval [0 °, 180 °] are chosen, 12 matched filters are created；
Green channel in the color retinal vascular images is done into convolutional calculation with 12 matched filters respectively, is obtained
To matched filtering response image, the matched filtering response image is normalized and is quantified as 256 grades of grayscale image, when described inclined
When being less than the threshold value of setting from degree σ, using obtained grayscale image as thin scale matched filtering response image；As the irrelevance σ
Not less than setting threshold value when, using obtained grayscale image as thick scale matched filtering response image.
3. according to the method described in claim 2, it is characterized in that, described utilize algorithm of region growing based on the seed point
Generate several line support areas, comprising:
Select a not used pixel as seed point from the sorted lists of pixel, by the gradient side of the seed point
To the initial angle θ as the line support area where the seed point to be generatedregion, will be in the seed neighborhood of a point
Not used and its gradient direction is with regional perspective θregionBetween pixel of the error between τ be added to the line and support
In region, the angle for calculating the line support area is updated according to updated pixel, wherein τ is angle threshold；
Above-mentioned treatment process is repeated, it is described until there is no qualified pixel to be added in the seed neighborhood of a point
In line support area, the corresponding minimum circumscribed rectangle in the line support area is extended.
4. method according to any one of claims 1 to 3, which is characterized in that the method further include:
The segmentation result of the thick vessel segment and the segmentation result of the thin vessel segment are merged using logic or operation, obtained
To the segmentation result of complete thick vessel segment and thin vessel segment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611185885.2A CN106651846B (en) | 2016-12-20 | 2016-12-20 | Segmentation method of retinal blood vessel image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611185885.2A CN106651846B (en) | 2016-12-20 | 2016-12-20 | Segmentation method of retinal blood vessel image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106651846A CN106651846A (en) | 2017-05-10 |
| CN106651846B true CN106651846B (en) | 2019-07-12 |
Family
ID=58834848
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611185885.2A Expired - Fee Related CN106651846B (en) | 2016-12-20 | 2016-12-20 | Segmentation method of retinal blood vessel image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106651846B (en) |
Families Citing this family (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107292890A (en) * | 2017-06-19 | 2017-10-24 | 北京理工大学 | A kind of medical image cutting method and device |
| CN107392922B (en) * | 2017-07-18 | 2019-12-03 | 东南大学 | Three-dimensional blood vessel segmentation method based on symmetry matched filter group and region growing |
| CN107492100B (en) * | 2017-09-30 | 2020-07-28 | 上海联影医疗科技有限公司 | Image processing method, device and equipment |
| CN108154513A (en) * | 2017-11-21 | 2018-06-12 | 中国人民解放军第三军医大学 | Cell based on two photon imaging data detects automatically and dividing method |
| CN110310323A (en) * | 2018-03-20 | 2019-10-08 | 天津工业大学 | The retinal blood vessels caliber measurement method being fitted based on Hessian matrix and dimensional Gaussian |
| CN108735270A (en) * | 2018-05-25 | 2018-11-02 | 杭州脉流科技有限公司 | Blood flow reserve score acquisition methods, device, system and computer storage media based on dimensionality reduction model |
| CN109325944A (en) * | 2018-09-13 | 2019-02-12 | 福建农林大学 | A kind of Segmentation Method of Retinal Blood Vessels based on support transformation and line detective operators |
| CN109410191B (en) * | 2018-10-18 | 2022-03-25 | 中南大学 | OCT (optical coherence tomography) image-based fundus blood vessel positioning method and anemia screening method thereof |
| CN109448003B (en) * | 2018-10-26 | 2021-03-02 | 强联智创（北京）科技有限公司 | Intracranial artery blood vessel image segmentation method and system |
| CN109544525A (en) * | 2018-11-15 | 2019-03-29 | 北京工业大学 | A kind of eyeground picture blood vessel recognition methods based on self-adapting window Model Matching |
| CN109671095B (en) * | 2018-12-19 | 2023-04-25 | 吉林大学 | Method and related device for separating metal objects in X-ray photo |
| CN109685814B (en) * | 2019-01-02 | 2020-10-23 | 兰州交通大学 | Full-automatic cholecystolithiasis ultrasonic image segmentation method based on MSPCNN |
| CN109978819B (en) * | 2019-01-22 | 2022-11-15 | 江南大学 | Method for segmenting retinal blood vessels based on low-scale blood vessel detection |
| CN110033861B (en) * | 2019-03-26 | 2021-03-23 | 西安交通大学 | Blood vessel and macular avascular region quantitative analysis method and system suitable for OCTA image |
| CN110517274B (en) * | 2019-08-30 | 2022-04-01 | 集美大学 | Image threshold segmentation method, terminal device and storage medium |
| CN110448267B (en) * | 2019-09-06 | 2021-05-25 | 重庆贝奥新视野医疗设备有限公司 | Multimode fundus dynamic imaging analysis system and method |
| CN111383191B (en) * | 2019-12-11 | 2024-03-08 | 北京深睿博联科技有限责任公司 | Image processing method and device for vascular fracture repair |
| CN111275668B (en) * | 2020-01-13 | 2023-09-15 | 浙江杜比医疗科技有限公司 | Method, system and device for extracting breast blood vessels of NIR (near infrared) image |
| CN111862114A (en) * | 2020-07-10 | 2020-10-30 | 温州医科大学 | Choroidal three-dimensional blood vessel imaging and quantitative analysis method and device based on optical coherence tomography system |
| CN112508888B (en) * | 2020-11-26 | 2023-12-19 | 中国科学院苏州生物医学工程技术研究所 | Brain artery rapid automatic segmentation method and system for medical image |
| CN112465852B (en) * | 2020-12-03 | 2024-01-30 | 国网山西省电力公司晋城供电公司 | Improved region growing method for infrared image segmentation of power equipment |
| CN112734784B (en) * | 2021-01-28 | 2024-07-19 | 依未科技(北京)有限公司 | Method, device, medium and equipment for determining fundus blood vessel boundary with high precision |
| CN112734828B (en) * | 2021-01-28 | 2023-02-24 | 依未科技(北京)有限公司 | Method, device, medium and equipment for determining center line of fundus blood vessel |
| CN113269756B (en) * | 2021-05-28 | 2024-02-27 | 长春大学 | Method and device for optimizing retinal vessel segmentation based on multi-scale matched filtering and particle swarm |
| CN114359205B (en) * | 2021-12-29 | 2022-11-01 | 推想医疗科技股份有限公司 | Head and neck blood vessel analysis method and device, storage medium and electronic equipment |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102999905A (en) * | 2012-11-15 | 2013-03-27 | 天津工业大学 | Automatic eye fundus image vessel detecting method based on PCNN (pulse coupled neural network) |
| CN103198321A (en) * | 2013-03-27 | 2013-07-10 | 中国科学院苏州生物医学工程技术研究所 | Retina image processing method and device |
| CN104036521A (en) * | 2014-07-03 | 2014-09-10 | 重庆大学 | Novel retina eye fundus image segmenting method |
| CN104318565A (en) * | 2014-10-24 | 2015-01-28 | 中南大学 | Interactive method for retinal vessel segmentation based on bidirectional region growing of constant-gradient distance |
| CN104899862A (en) * | 2015-04-01 | 2015-09-09 | 武汉工程大学 | Retinal vessel segmentation algorithm based on global or local threshold |
| CN105761258A (en) * | 2016-02-06 | 2016-07-13 | 上海市第人民医院 | Retinal fundus image bleeding detection method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8351669B2 (en) * | 2011-02-01 | 2013-01-08 | Universidade Da Coruna-Otri | Method, apparatus, and system for retinal image analysis |
-
2016
- 2016-12-20 CN CN201611185885.2A patent/CN106651846B/en not_active Expired - Fee Related
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102999905A (en) * | 2012-11-15 | 2013-03-27 | 天津工业大学 | Automatic eye fundus image vessel detecting method based on PCNN (pulse coupled neural network) |
| CN103198321A (en) * | 2013-03-27 | 2013-07-10 | 中国科学院苏州生物医学工程技术研究所 | Retina image processing method and device |
| CN104036521A (en) * | 2014-07-03 | 2014-09-10 | 重庆大学 | Novel retina eye fundus image segmenting method |
| CN104318565A (en) * | 2014-10-24 | 2015-01-28 | 中南大学 | Interactive method for retinal vessel segmentation based on bidirectional region growing of constant-gradient distance |
| CN104899862A (en) * | 2015-04-01 | 2015-09-09 | 武汉工程大学 | Retinal vessel segmentation algorithm based on global or local threshold |
| CN105761258A (en) * | 2016-02-06 | 2016-07-13 | 上海市第人民医院 | Retinal fundus image bleeding detection method |
Non-Patent Citations (3)
| Title |
|---|
| A Cross-Modality Learning Approach for Vessel Segmentation in Retinal Images;Qiaoliang Li.et.;《IEEE TRANSACTIONS ON MEDICAL IMAGING》;20160131;第35卷(第1期);第109-118页 * |
| Retinal vessel extraction by matched filter with first-order derivative of Gaussian;Bob Zhang.et.;《Computers in Biology and Medicine》;20100430;第40卷(第4期);第438-445页 * |
| 基于多尺度2D Gabor小波的视网膜血管自动分割;王晓红等;《自动化学报》;20150531;第41卷(第5期);第970-980页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106651846A (en) | 2017-05-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106651846B (en) | Segmentation method of retinal blood vessel image | |
| Kumar et al. | An automated early diabetic retinopathy detection through improved blood vessel and optic disc segmentation | |
| Yang et al. | An automatic hybrid method for retinal blood vessel extraction | |
| US8463065B2 (en) | Linear feature detection method and apparatus | |
| US11944499B2 (en) | Tracking contrast elements in ultrasound imaging | |
| Kumar et al. | Image processing in diabetic related causes | |
| Kaushik et al. | Medical image segmentation using genetic algorithm | |
| Purnama et al. | Follicle detection on the usg images to support determination of polycystic ovary syndrome | |
| Monisha et al. | Computerized Diagnosis of Diabetic Retinopathy based on Deep Learning Techniques | |
| JP2003502129A (en) | Computer aided detection method for lumps and clustered microcalcifications | |
| Li et al. | An automated method using hessian matrix and random walks for retinal blood vessel segmentation | |
| HongQing | Segmentation of blood vessels in retinal images using 2D entropies of gray level-gradient cooccurrence matrix | |
| Belhadi et al. | Automated retinal vessel segmentation using entropic thresholding based spatial correlation histogram of gray level images | |
| Sreedhanya et al. | An automatic brain tumor detection and segmentation using hybrid method | |
| Azegrouz et al. | Max-min central vein detection in retinal fundus images | |
| Pécot et al. | Non parametric cell nuclei segmentation based on a tracking over depth from 3d fluorescence confocal images | |
| Köse | A simple hybrid method for segmenting vessel structures in retinal fundus images | |
| CN110021016A (en) | A kind of calcification detection method | |
| Rohman et al. | New techniques for segmentation and extraction retinal blood vessels | |
| Lupascu et al. | A multiscale approach to automatic and unsupervised retinal vessel segmentation using Self-Organizing Maps | |
| CN118570556B (en) | Mammary tumor classification method based on multi-sequence MRI feature enhancement | |
| Liu et al. | Identification and classification of cells in multispectral microscopy images of lymph nodes | |
| Jichkar et al. | Detection of Retinal Blood Vessels for Diagnosing Diseases | |
| EP3910587A1 (en) | Computer-implemented method, system and computer program product for determining a vascular function of a perfusion imaging sequence | |
| Xu et al. | Using PSO to improve dynamic programming based algorithm for breast mass segmentation |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190712 |