LU102573B1 - Clustering Algorithm-based Multi-parameter Cumulative Calculation Method for Lower Limb Vascular Calcification Indexes - Google Patents
Clustering Algorithm-based Multi-parameter Cumulative Calculation Method for Lower Limb Vascular Calcification Indexes Download PDFInfo
- Publication number
- LU102573B1 LU102573B1 LU102573A LU102573A LU102573B1 LU 102573 B1 LU102573 B1 LU 102573B1 LU 102573 A LU102573 A LU 102573A LU 102573 A LU102573 A LU 102573A LU 102573 B1 LU102573 B1 LU 102573B1
- Authority
- LU
- Luxembourg
- Prior art keywords
- super
- image
- cumulative
- points
- pixel
- Prior art date
Links
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
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/12—Edge-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/181—Segmentation; Edge detection involving edge growing; involving edge linking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30101—Blood vessel; Artery; Vein; Vascular
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical 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)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The invention discloses a clustering algorithm-based multi-parameter cumulative calculation method for lower limb vascular calcification indexes, comprising the steps: firstly carrying out super-pixel segmentation of a CT image, and enabling calcified spots in the CT image to be segmented in each super-pixel region; after the super-pixel segmentation is accomplished, extracting a brightness characteristic value of a super-pixel region where the calcified spots are located by using a Lab color space, and performing edge detection and contour extraction on the calcified spots in the image; then fitting the calcified spots in the image by using a segmented ellipse, and extracting the area of the calcified spots after optimizing an ellipse contour. According to the clustering algorithm-based multi-parameter cumulative calculation method for lower limb vascular calcification indexes, the errors of experience judgment are effectively reduced.
Description
]
DESCRIPTION Clustering Algorithm-based Multi-parameter Cumulative Calculation Method for Lower Limb Vascular Calcification Indexes
TECHNICAL FIELD The invention belongs to the technical field of medical image processing, and specifically relates to a clustering algorithm-based multi-parameter cumulative calculation method for lower limb vascular calcification indexes.
BACKGROUND Diabetic lower extremity arteriosclerosis obliterans, important risk factors of which are diabetes, hypertension, lipid metabolism disorders, and smoking, is a more serious peripheral arterial vascular disease in which diabetes has developed to a certain degree. Due to long-term stimulation of a hyperglycemic factor, diabetic patients often have a more serious degree of peripheral arterial disease in comparison with patients with hypertension, and patients with lower extremity arterial disease are more likely to have higher disability and mortality. Patients with mild conditions may have numbness and chills in the lower limbs while patients with moderate and severe conditions are likely to have intermittent rest pain, and even necrosis that leads to amputation and disability. Due to the complexity of diabetes and the long-term course of the disease, diabetic lower extremity arteriosclerosis obliterans tends to have more extensive and severe lesions than lower extremity arteriosclerosis obliterans caused by other reasons such as more plaques, especially small vascular lesions below à Knee being most severs, thus bringing certain difficulty to civics treatment due toits complexity.
At present, the examination techniques, which also have shurtcomings, for lower extremity arterss include ankle-brachial index measurement, arigriography, Color Doppler, CT anglography, magneïn resonance angiography and other methods The anklebrachial index messurement cannot judgs the degres and nature of vascular stenosis, and there is à false negative rats when calcHioation is mors serous. Aderiography is an invasive examination with high cost and many complications. The examination results of Color Doppler are greatly affected by factors such as the operatore proficiency, and the display of deep blood vessels and adiscent bone blood vessels 8 pour. The magnetic rmsonance angiography, although used mors and more widely, Is low in spatial resolution, has relatively large deviation in diagnosis of small blood vessels, and cannot fully meet the clinical needs. The LT image technology, although used More and mors widely, is greatly affsciad ry human factors on judgement results when CÉTICIENE mostly assess patients” corudiions from imaging results based on persans! axpartance, and therafore, it is necessary Lo design 3 more effective method for the CT image, which can ouiput relatively accurate lover mb vascular caicification degres of and affectivaly prevent human factors from affecting the jwdgment results through processing of the CT image.
In addition, unlike the distribution of calcified plaques in other arteries of a human body, the branches of lower extremity arteries are more refined and complicated. It is the key point and the difficulty in research at present to correctly obtain vascular caicification indexes of lower limbs as diagnostic accuracy diabetic feet facing amputation is often affected by calcification degree of the calcified plaques in the blood vessels of lower limbs.
SUMMARY In order to solve the problem that there are no effective calculation methods for lower limb vascular calcification indexes in the prior art, the invention discloses a clustering algorithm-based multi-parameter cumulative calculation method for lower limb vascular calcification indexes, which gives quantitative values of the lower limb vascular calcification indexes by combining a CT brightness characteristic value obtained by CT image processing of the blood vessels of lower limbs and a calcification area with corresponding cumulative correction coefficient obtained by vascular fluid mechanics study. The values of the lower limb vascular calcification indexes can represent calcification degree of the blood vesseis of lower limbs to a great extent, and provide data basis for the risk that diabetic feet face amputation subsequentially.
The technical scheme adopted by the invention is as follows: step 1, acquiring a CT image of the blood vessels of lower limbs to be analyzed,
step 2, using a linear iterative clustering algorithm to evenly segment the calcified spots in the CT image into each super-pixel region; step 3, after accomplishing the super-pixel segmentation, extracting a brightness characteristic value of a super-pixel region where the calcified spots are located by using a Lab color space; step 4, performing edge detection and contour extraction on the calcified spots in the CT image, fitting the calcified spots in the processed image by using a segmented ellipse and optimizing to obtain a radius of the calcified spots, thereby calculating the area of the calcified spots; and step 5, obtaining a judgment value of the calcification degree in the CT image according to Cal=k.p.S, where ” is a CT brightness characteristic value, S is an area of the calcified spots, and k is a cumulative correction coefficient.
Further, the step 2 of performing super-pixel segmentation on the CT image is as follows: step 2.1, performing even super-pixel segmentation on the acquired original CT image of the blood vessels of the lower limbs, setting X pixels in the original CT image of the blood vessels of the lower limbs, segmenting the original CT image of the blood vessels of the lower limbs into K regions, where
X each super-pixel has K pixels;
step 2.2, presetting an interval between an initial clustering center C and an initial clustering center C; step 2.3, searching pixels close to C in the field of clustering center C based on a Euclidean distance, and classifying the pixels into one category: step 2.4, calculating an average eigenvector value of all pixels in the K super-pixel regions, performing next clustering based on the average eigenvector value, iteratively updating the clustering center, and iterating again until the end of the iteration; and step 2.5, segmenting the iterated super pixels to obtain the super-pixel regions.
Further, the step 3 of extracting the brightness characteristic value of the super-pixel region where the calcified spots are located is as follows: step 3.1, extracting a brightness channel L in Lab and representing L=] or 1] 16 brightness characteristic fo of the CT image of the blood vessels of the lower limbs to obtain a brightness image Lo of the CT image of the blood vessels of the lower limbs, where Y is an intermediate variable, Yo is a gray value of white defined by the CIE standard, and f is a correction function;
and step 3.2, based on a brightness map Lo, extracting the brightness characteristic value of the super-pixel region where the calcified spots are located.
Further, the step 3.2 of extracting the brightness characteristic value is as follows: step 3.2.1, using Gaussian-filtering 1/2 down-sampling to process Lo to obtain an image Li, Li=subsample {(Ipfilter (Lo)), where subsample () is a down-sampling function and Ipfilter () is a frequency domain filter function; step 3.2.2, extracting the maximum brightness pixel point A *¥ in the super-pixel region subjected to Gaussian filtering, and obtaining a sum P (x)= f(xy) Fn of the gray values of all pixels in the super-pixel region, where (x, y) are pixel coordinates, yen) x (x, %,) and yı and yz are coordinate values in the y-axis direction in the super-pixel region, x1 and xz are coordinate values in the x-axis direction in the super-pixel region, f(x,y) is pixel values at (x,y), Py(x) is a cumulative sum of the gray values of the column vector pixels at the x, and a sum of grey values of all pixels in the entire P=Y 1x) super-pixel region is obtained by x and step 3.2.3, obtaining a cumulative sum of the gray values of all pixels in the entire super-pixel region as the CT brightness value.
Further, the process of extracting the calcified spots by edge detection and contour extraction is as follows: step 4.1.1, using a Gaussian filter to preprocess the image; step 4.1.2, using a sobel operator to calculate a gradient size and a direction of each pixel point in the filtered image; step 4.1.3, selecting edge points based on the gradient intensity comparison of the pixels, keeping the edge points while the gradient intensity of certain pixel is greater than that of another two pixels in a positive and negative gradient direction, otherwise, suppressing the pixel; and step 4.1.4, comparing the edge points obtained in the previous step with an upper threshold, and then screening the edge points; if the upper threshold is less than the edge points, keeping the point and setting the changed point as the first edge point; then searching whether or not a neighboring point of this point is less than the upper threshold, repeating this process and connecting all the points greater than the upper threshold.
Further, the method of fitting by the segmented ellipse is as follows: step 4.2.1, randomly segmenting the obtained contour into n segments; step 4.2.2, randomly selecting 12 non-repetitive points in each segment of the contour, and using the least squares method to fit n candidate ellipses; step 4.2.3, setting a judgment threshold value lo, comparing a distance li between the point (xi, yi) and the candidate ellipse contour with a judgment threshold value lo; if li is greater than lo, discarding the changed point and not recording; if li is smaller than or equal to lo, keeping the changed point, recording as one, and summarizing the relevant parameters of the point to obtain a data set Vi=(xic,Yic,Ai, bi,8i,ni,s), where the circle center of the candidate ellipse is (Xi, Yic), à semi-major axis is ai, a semi-minor axis is bi, a rotation angle is 6i, si represents a serial number of each segment, and ni is the number of contour segments; and repeating the above comparison process until all points on the candidate ellipse contour are compared, summing up all the kept data set of the M points to obtain V={Vi=(xic,Yic,ai, Di, 0i,ni,si)li=1,2,...M }, where the one with the most votes is the fitted result.
Further, the ellipse contour is obtained by an active contour model to obtain the area of the calcified spots.
step 4.3.1, using a snake model to give a 2D parameter closed curve near the region of interest, and by minimizing the energy functional, deforming the closed curve in the image and continuously approximating the target contour, receiving the final evolution results as the target contour, and expressing a contour curve energy function as follows: | En (30) = | BY + EYE 0 0 Where, Esnake (V(S)) is curve energy, v(s) is a parameter equation of snake contour, Eint is internal energy of the curve, which determines the smoothness and continuity of the curve; Eex is energy given to the curve by the outside,
which makes the curve move towards a characteristic direction of the target, and s is an independent variable describing the boundary; and step 4.3.2, using the least-square circle fitting method to re-fit the circle, and getting the center of the circle through a weighting function of the coordinates of the edge points on all the circles, that is, the centers (X, Y) of 1 £ 1 € X= —>x Y=— dy the calcified spots, where Nia , N and the diameters i=l of the calcified spots are calculated, xi and vi respectively represent the coordinates of certain point on the contours of the calcified spots, and N is the number of points on the contours of the calcified spots; and finally getting the area of the calcified spots.
Further, the cumulative correction coefficient k=kn*kw*ko*kp, kn is a cumulative score of stenosis; kw is a cumulative score of wall shear stress w of the lower extremity arteries; ko is a cumulative score of an oscillating shear index o of the lower extremity arteries; kp is a cumulative score of wall shear stress p of the lower extremity arteries; the cumulative calculation standard for lower extremity arterial stenosis is as follows: according to the lumen area of the lower extremity arteries, it is divided into four grades: |. under lumen diameter reduction of 1%-25%, cumulative stenosis score knis 1 point. Il. under lumen diameter reduction of 25%-50%, stenosis cumulative score kn is 2 points, lll. under lumen diameter reduction 51%-75%, stenosis cumulative score kn is 3 points, IV. under lumen diameter reduction 76%-100%, stenosis cumulative score kn is 4 points, where the method of collecting stenosis data is as follows: collecting the maximum value of the vascular stenosis hmax and the average value h of the stenosis, in the segment with calcified plaques, of the lower extremity artery, and determining the value “of vascular stenosis of this segment according to h=ah,, +bh , where a and b are constant coefficients.
The cumulative calculation standard for wall shear stress w of lower limb arteries is as follows: wall shear stress in an arterial system is generally (10-70) dynes/cm?, and when wall shear stress w is (0-4) dynes/cm?, the cumulative score kw is 3 points, when the wall shear stress w is (5-10) dynes/cm?, the cumulative score kw is 2 points, and when the wall shear stress w is (11-70) dynes/cm?, the cumulative score kw is 1 point.
The accumulative calculation standard for the oscillating shear index o of the lower limb arteries is as follows: the normal range of the index is 0-0.5, when the oscillating shear index is lower than 0.2, the cumulative score ko is 3 points, the cumulative score Ko is 2 points when the oscillating shear index is between 0.2 and 0.3, and when the oscillating shear index is between 0.3 and
0.5, the cumulative score ko is 1 point.
The accumulative calculation standard for wall shear stress p of lower limb arteries is as follows: a normal range of indicators is systolic blood pressure of
90-140mmHg, diastolic blood pressure of 60-90mmHg; when the systolic blood pressure is lower than 30mmHg and the diastolic blood pressure is lower than 60mmHg, the cumulative score kp is 3 points; when the systolic blood pressure is between 90-140mmHg and the diastolic blood pressure is between 60-90mmHg, the cumulative score kp is 2 points,; and when the systolic blood pressure is higher than 140mmHg and the diastolic blood pressure is higher than 90mmHg, the cumulative score kp is1 point.
The invention has the following beneficial effects: Firstly, the clustering algorithm-based multi-parameter cumulative calculation method for lower limb vascular calcification indexes disclosed by the invention firstly performs super-pixel segmentation on the CT image of the blood vessels of the lower limbs in a process of processing the CT image of the blood vessels of the lower limbs, so that on one hand, the pixels can be aggregated together to form multiple sub-region blocks with regular shapes and consistent local structures, and thus, overall expression of image local factures and structural information is realized, excessive data are avoided, processing speed is increased, and data dimensions are reduced by super pixels; on the other hand, the linear iterative clustering algorithm is used to calculate the average value of the features in the super-pixel region to replace the pixel value in the region, which can keep effective information to the greatest extent and reduce noises.
Secondly, in the process of extracting the brightness characteristic value of the super-pixel region where the calcified spots are located by using a Lab color space, a low-pass filter and sub-sampling operation are firstly used layer by layer on the original image Lo to obtain an spatial-scale-transform brightness intensity map of the original image Lo, which can enhance the edge of the salient area of the image; then, the sum of the gray values of all pixels in the super pixel area is calculated according to the pixel point with the maximum brightness in the super pixel region, and the CT brightness value can be obtained from the Lab color space by using pixel gray value integral.
Thirdly, when the calcified area in the CT image is extracted, a Canny operator is used to perform edge detection and contour extraction on the calcified spots in the image, and a Gaussian filter is used to preprocess the image to reduce the effect of noises; non-maximum suppression is used to make the edge have an accurate response, and the accuracy of the target is improved by edge detection on a correct position. Hysteresis threshold processing detection is used to connect edge points, and remove false edges, so that edge positioning precision is improved.
Fourthly, the present adopts an active contour model to optimize the ellipse contour, that is, the snake model is used to deform the contour under the action of internal and external forces, and the external energy attracts the active contour to continuously approach the target contour, and is finally evolved to be received as the target contour, where the contour which is evolved through the snake model curve and is refitted is closer to the real contour.
And fifthly, the clustering algorithm-based multi-parameter cumulative calculation method for lower limb vascular calcification indexes gives quantitative values of the lower limb vascular calcification indexes by combining a CT brightness characteristic value obtained by CT image processing of the blood vessels of lower limbs and a calcification area with corresponding cumulative correction coefficient obtained by vascular fluid mechanics study. The values of the lower limb vascular calcification indexes can represent calcification degree of the blood vessels of lower limbs to a great extent, provide data basis for the risk that diabetic feet face amputation subsequentially, and effectively reduce errors caused by experienced judgment. Besides, the calcification degree is obtained by processing the CT image, the influence of human judgment factor is avoided and accuracy is improved.
BRIEF DESCRIPTION OF THE FIGURES Fig. 1 is flow chart of the method of the invention.
Fig. 2 is flow chart of dividing the super pixel region.
Fig. 3 is a CT image of the lower limbs of a diabetic patient.
In order to make the objective, technical scheme and advantages of the invention clear, the invention is further illustrated in detail in combination with the accompanying drawings and embodiments hereinafter. It should be understood that specific embodiments described herein are only used for explaining the invention, instead of limiting the invention.
Step 1, a CT image of the blood vessels of lower limbs of a diabetic patient was acquired by detecting instruments such as a CT to obtain a CT image of the blood vessels of lower limbs to be analyzed, step 2, a linear iterative clustering algorithm (SLIC) was used to evenly segment the calcified spots in the CT image into each super-pixel region as follows: step 2.1, X pixels in the original CT image of the blood vessels of the lower limbs were set, K initial super pixels were set, and the original CT image was
X evenly segmented, where each cut each super-pixel region had À pixels: step 2.2, an interval between an initial clustering center C and an initial clustering center C was preset; step 2.3, pixels close to C in the field of clustering center C were searched based on a Euclidean distance, and the pixels were classified into one category until all pixel points in the field where each clustering center C was located were segmented;
step 2.4, an average eigenvector value of all pixels in the K super-pixel regions was calculated, next clustering was performed based on the average eigenvector value, the clustering center was iteratively updated and iterated again until the end of the iteration; step 2.5, the iterated super pixels were segmented to obtain the super-pixel regions.
Further, the step 3 of extracting the brightness characteristic value of the super-pixel region where the calcified spots were located was as follows: step 3.1, a brightness channel L in Lab was extracted and brightness L=1 or 2 )-1s characteristic Ka of the CT image of the blood vessels of the lower limbs was represented to obtain a brightness image Lo of the CT image of the blood vessels of the lower limbs, where Y was an intermediate variable, Yo was a gray value of white defined by the CIE standard, and f was a correction function; and step 3.2, based on a brightness map Lo of the L channel, the brightness characteristic value of the super-pixel region where the calcified spots were located was extracted as follows: step 3.2.1, brightness map Lo of the L channel was read, Lo was used as the original image, and the image Li was got through Gaussian filtering 1/2 down-sampling, L1 being equal to subsample(ipfilter(Lo)), where subsample was a down-sampling function; Lpfilter() was a frequency domain filter function; an image Li with Lo intensity could be obtained after frequency domain filtering and down-sampling operations were performed on Lo, and the edges of the salient areas of the processed image were effectively enhanced; step 3.2.2, the maximum brightness pixel point A ®¥ in the super-pixel region in the brightness-enhanced image Li was extracted, and a sum Ja ; P(x)= 2 (x) = of the gray values of ali pixels in the super-pixel region was obtained, where (x, y) were pixel coordinates, YENI), xen, x) and y1 and yı were coordinate values in the y-axis direction in the super-pixel region, x1 and x2 were coordinate values in the x-axis direction in the super-pixel region, f(x,y) was pixel values at (x,y), Py(x) was a cumulative sum of the gray values of the column vector pixels at the x, and a sum of grey values of all P=3P (x) pixels in the entire super-pixel region was obtained by = ; and step 3.23, a cumulative sum of the gray values of all pixels in the entire super-pixel region was obtained as the CT brightness value.
step 4, a calcified area in the CT image was extracted, and calcified spots distributed at the lower limbs of the patients could be shown as spots with unsmooth peripheries from the CT image, and thus, the process of extracting the calcified area from the image was as follows: step 4.1, the process of extracting the calcified spots by edge detection and contour extraction was as follows:
step 4.1.1, A Gaussian filter was used to preprocess the image for smoothing the image and filtering noises; step 4.1.2, a sobel operator was used to calculate a gradient size and a direction of each pixel point in the filtered image; step 4.1.3, edge points were selected based on the gradient intensity comparison of the pixels, the edge points while the gradient intensity of certain pixel is greater than that of another two pixels in a positive and negative gradient direction were kept, otherwise, the pixel point was suppressed to 0; and the edge points could generate accurate response through the selection process, so that accuracy of extracting the edge points was improved; and step 4.1.4, the edge points obtained in the previous step were compared with a set upper threshold, and then the edge points were screened; if the upper threshold is less than the edge points, the point was kept and the changed point was set as the first edge point; then whether or not a neighboring point of this point is less than the upper threshold was searched, this process was repeated and all the points greater than the upper threshold were connected; and false edges were removed, so that edge positioning precision was improved.
Step 4.2, the method of fitting by the segmented ellipse was follows:
step 4.2.1, the obtained contour obtained in the previous step was randomly segmented into n segments, nels, 12] where n was an even number within the range;
step 4.2.2, 12 non-repetitive points in each segment of the contour were randomly selected, and the least squares method was used to fit n candidate ellipses;
step 4.2.3, a judgment threshold value lo was set, a distance li between the point (xi, yi) and the candidate ellipse contour was compared with a judgment threshold value lo; if li was greater than lo, the changed point was discarded and not recorded; if li was smaller than or equal to lo, the changed point was kept and recorded as one, and the relevant parameters of the point to obtain a data set Vi=(xic,Yic,Ai,Di,Bi,Ni,Si) were summarized, where the circle center of the candidate ellipse was (xic, yi, a semi-major axis was ai, a semi-minor axis was bi, a rotation angle was 8i, sirepresented a serial number of each segment, and ni was the number of contour segments; and the above comparison process was repeated until all points on the candidate ellipse contour were compared, all the kept data set of the M points was summed up to obtain V={Vi=(xic,Vic,Ai, bi, 8i,ni,si)[i=1,2,..,M }, where the one with the most votes was determined as the candidate circle.
The invention used the segmented ellipse fitting to effectively reduce the probability of fitting an incorrect ellipse and improved the accuracy.
step 4.3, an ellipse contour was optimized by an active contour model to finally obtain the area of the calcified spots:
step 4.3.1, a snake model was used to give a 2D parameter closed curve near the region of interest, and by minimizing the energy functional, the closed curve was deformed in the image and continuously approximate the target contour, the final evolution results were received as the target contour, and a contour curve energy function was expressed as follows:
| Ba v(s))ds = | be (5) + E, (v(s))ds
0 0
Where, Esnake (v(S)) was curve energy, v(s) was a parameter equation of snake contour, Eint was internal energy of the curve, which determined the smoothness and continuity of the curve; Eex was energy given to the curve by the outside, which made the curve move towards a characteristic direction of the target, and s was an independent variable describing the boundary;
and step 4.3.2, the least-square circle fitting method was used to re-fit the circle, and the center of the circle was got through a weighting function of the coordinates of the edge points on all the circles, that is, the centers (X, Y) of
1 € 1 € X=—>%x y==S5y the calcified spots, where Nas , N57 and the diameters
24 > 2 Dep 2x X) +(»x -Y) | of the calcified spots was calculated, xi and vy;
respectively represented the coordinates of certain point on the contours of the calcified spots, and N was the number of points on the contours of the calcified spots; and the area of the calcified spots was finally got.
Step 5, the CT brightness value p and the area S of the calcification spots obtained in the above steps 3 and 4 were substituted into Cal =k p 8 to obtain a judgment value of the calcification degree in the CT image, where k was the cumulative correction coefficient, and the calcification indexes not only related to the areas S of the calcified spots in the CT image, the brightness of the calcified spots, and the stenosis, but also closely related to the wall shear stress w, the oscillating shear index o, and the wall pressure p. The blood vessels which were 4cm long in the region where the calcified spots were located were intercepted. The cumulative correction coefficient k was used to correct the calcification degree, the dynamics of blood flow changed significantly in the stenosis section after stenosis appeared in the arterial lumen as types of calcification in the lower extremity arteries included intimal calcification and media calcification, and according to the calcified spots in the CT image, it was impossible to directly distinguish the intimal calcification or media calcification, but intimal calcification could easily cause stenosis of the arterial lumen. Therefore, in the invention by increasing the cumulative correction coefficient k to comprehensively calculate the CT brightness value p and the area S of the calcified spots, so that a more accurate judgment value of the calcification degree was obtained, and the judgment value of the calcification index in the CT image was obtained according to Cal = k.p. S;
where, the cumulative calculation of the area S of calcified plaques of the lower extremity artery could be directly obtained by fitting the calcified spots in the image processed using the segmented ellipse and performing optimization processing to obtain the radius of the calcified spots, and then calculating the area of the calcified spots, or the cumulative calculation was directly given by the scoring standard: the size of the calcified plaque area on the anterior and posterior walls of the lower extremity arteries was divided into four grades: |, when there was no calcification, scored 0; Il, when calcification range was less than 1/3 of the arterial wall length, scored 1; Ill. When calcification range was 1/3 to 2/3 of the arterial wall length, scored 2; IV. When calcification range was greater than 2/3 of the arterial wall length, scored 3. The blood vessels of the lower limbs were divided into the upper and lower sections by taking the middle of shanks as boundary lines, including: proximal posterior tibial artery, distal tibial artery, proximal anterior tibial artery, distal tibial artery, proximal peroneal artery, and distal peroneal artery end.
The brightness value p of the super pixel region where the calcification spots were located can be directly obtained by extracting the CT brightness characteristic value of the super pixel region where the calcification spots were located through the Lab color space in the clustering algorithm.
The above cumulative correction coefficient could be obtained according to the following calculation formula: k=kn*kw*ko*kp, where kn was the cumulative score of stenosis; kw was the cumulative score of the wall shear stress w of the lower extremity arteries; ko was the cumulative score of the oscillating shear index o of the lower extremity arteries; kp was the cumulative score of wall shear stress p of the lower extremity arteries; and the calculation standard of specific cumulative correction coefficient k was as follows:
the cumulative calculation standard for lower extremity arterial stenosis was as follows: according to the lumen area of the lower extremity arteries, it was divided into four grades: |. under lumen diameter reduction of 25%-50%, cumulative stenosis score kn was 1 point; Il. under lumen diameter reduction of 25%-50%, stenosis cumulative score kn was 2 points; Ill. under lumen diameter reduction 51%-75%, stenosis cumulative score kh was 3 points; IV. under lumen diameter reduction 76%-100%, stenosis cumulative score kn was 4 points, where the method of collecting stenosis data was as follows: collecting the maximum value of the vascular stenosis hmax and the average value # of the stenosis, in the segment with calcified plagues, of the lower extremity artery, and determining the value #of vascular stenosis of this segment according to h = ah, +bh , where a and b were constant coefficients.
The cumulative calculation standard for wall shear stress w of lower limb arteries was as follows: wall shear stress in an arterial system was generally (10-70) dynes/cm?, and when wall shear stress w was (0-4) dynes/cm?, the cumulative score kw was 3 points, when the wall shear stress w was (5-10) dynes/cm”, the cumulative score kw was 2 points, and when the wall shear stress w was (11-70) dynes/cm?, the cumulative score kw was 1 point.
The accumulative calculation standard for the oscillating shear index o of the lower limb arteries was as follows: the normal range of the index was 0-0.5, when the oscillating shear index was lower than 0.2, the cumulative score ko was 3 points, and when the oscillating shear index was between 0.2 and 0.3, the cumulative score ko was 2 points, the cumulative score ko was 1 point when the oscillating shear index was between 0.3 and 0.5.
The accumulative calculation standard for wall shear stress p of lower limb arteries was as follows: a normal range of indicators was systolic blood pressure of 90-140mmHg, diastolic blood pressure of 60-90mmHg; when the systolic blood pressure was lower than 90mmHg and the diastolic blood pressure was lower than 60mmHg, the cumulative score kp was 3 points; when the systolic blood pressure was between 90-140mmHg and the diastolic blood pressure was between 60-90mmHg, the cumulative score kp was 2 points; and when the systolic blood pressure was higher than 140mmHg and the diastolic blood pressure was higher than 90mmHg, the cumulative score kp was 1 point.
The clustering algorithm-based multi-parameter cumulative calculation method for lower limb vascular calcification indexes of the invention does not directly aim to obtain the diagnosis results or health status of diabetic foot disease, but only obtains information as the intermediate result from the living human body, which provides intermediate data support for diagnosis of the diabetic foot disease, and does not belong to the category of disease diagnosis.
The embodiment discussed above is merely for describing the technical concepts and features of the invention, the objectives are that those skilled in this art could understand the content of the invention and implement therefrom, limitation to the patent scope of the invention cannot be made only by this embodiment, that is to say, any equivalent variations or modifications in accordance with the spirit disclosed by the invention shall be contemplated as being within the patent scope of the invention.
Claims (8)
1. A clustering algorithm-based multi-parameter cumulative calculation method for lower limb vascular calcification indexes, characterized by comprising the steps: 1, acquiring a CT image of the blood vessels of lower limbs to be analyzed; 2, using a linear iterative clustering algorithm to evenly segment the calcified spots in the CT image into each super-pixel region; 3, after accomplishing the super-pixel segmentation, extracting a brightness characteristic value of a super-pixel region where the calcified spots are located by using a Lab color space; 4, performing edge detection and contour extraction on the calcified spots in the CT image, fitting the calcified spots in the processed image by using a segmented ellipse and optimizing to obtain a radius of the calcified spots, thereby calculating the area of the calcified spots; and 5, obtaining a judgment value of the calcification degree in the CT image according to Cal = k.p.S, where # is a CT brightness characteristic value, S is an area of the calcified spots, and k is a cumulative correction coefficient.
2. The clustering algorithm-based multi-parameter cumulative calculation method for lower limb vascular calcification indexes according to claim 1,
wherein the step 2 of performing super-pixel segmentation on the CT image is as follows: step 2.1, performing even super-pixel segmentation on the acquired original CT image of the blood vessels of the lower limbs, setting X pixels in the original CT image of the blood vessels of the lower limbs, segmenting the original CT image of the blood vessels of the lower limbs into K regions,
A wherein each super-pixel has K pixels; step 2.2, presetting an interval between an initial clustering center C and an initial clustering center C; step 2.3, searching pixels close to C in the field of clustering center C based on a Euclidean distance, and classifying the pixels into one category; step 2.4, calculating an average eigenvector value of all pixels in the K super-pixel regions, performing next clustering based on the average eigenvector value, iteratively updating the clustering center, and iterating again until the end of the iteration; and step 2.5, segmenting the iterated super pixels to obtain the super-pixel regions.
3. The clustering algorithm-based multi-parameter cumulative calculation method for lower limb vascular calcification indexes according to claim 1, wherein the step 3 of extracting the brightness characteristic value of the super-pixel region where the calcified spots are located is as follows:
step 3.1, extracting a brightness channel L in Lab and representing L=167(£)-10 brightness characteristic Ö of the CT image of the blood vessels of the lower limbs to obtain a brightness image Lo of the CT image of the blood vessels of the lower limbs, where Y is an intermediate variable, Yo is a gray value of white defined by the CIE standard, and f is a correction function; and step 3.2, based on a brightness map Lo of the brightness channel L, extracting the brightness characteristic value of the super-pixel region where the calcified spots are located.
4. The clustering algorithm-based multi-parameter cumulative calculation method for lower limb vascular calcification indexes according to claim 3, wherein the step 3.2 of extracting the brightness characteristic value is as follows: step 3.2.1, using Gaussian-filtering 1/2 down-sampling to process Lo to obtain an image Li, Li=subsample (Ipfilter (Lo)), wherein subsample () is a down-sampling function and Ipfilter () is a frequency domain filter function; step 3.2.2, extracting the maximum brightness pixel point A“ in the super-pixel region subjected to Gaussian filtering, and obtaining a sum P(«)= > f(xy) ven of the gray values of all pixels in the super-pixel region, wherein (x, y) are pixel coordinates, Je Va), X €(x,%,) and y1 and yz are coordinate values in the y-axis direction in the super-pixel region, x: and x2 are coordinate values in the x-axis direction in the super-pixel region, f(x,y) is pixel values at (x,y), Py(x) is a cumulative sum of the gray values of the column vector pixels at the x, and a sum of grey values of all pixels in the entire P=3 P(x) super-pixel region is obtained by U ; and step 3.2.3, obtaining a cumulative sum of the gray values of all pixels in the entire super-pixel region as the CT brightness value.
5. The clustering algorithm-based multi-parameter cumulative calculation method for lower limb vascular calcification indexes according to claim 1, wherein the process of extracting the calcified spots by edge detection and contour extraction is as follows: step 4.1.1, using a Gaussian filter to preprocess the image; step 4.1.2, using a sobel operator to calculate a gradient size and a direction of each pixel point in the filtered image step 4.1.3, selecting edge points based on the gradient intensity comparison of the pixels, keeping the edge points while the gradient intensity of certain pixel is greater than that of another two pixels in a positive and negative gradient direction, otherwise, suppressing the pixel, and step 4.1.4, comparing the edge points obtained in the previous step with an upper threshold, and then screening the edge points; if the upper threshold is less than the edge points, keeping the point and setting the changed point as the first edge point; then searching whether or not a neighboring point of this point is less than the upper threshold, repeating this process and connecting all the points greater than the upper threshold.
6. The clustering algorithm-based multi-parameter cumulative calculation method for lower limb vascular calcification indexes according to claim 5, wherein the method of fitting by the segmented ellipse is as follows: step 4.2.1, randomly segmenting the obtained contour into n segments; step 4.2.2, randomly selecting 12 non-repetitive points in each segment of the contour, and using the least squares method to fit n candidate ellipses; step 4.2.3, setting a judgment threshold value lo, comparing a distance li between the point (xi, yi) and the candidate ellipse contour with a judgment threshold value lo; if li is greater than lo, discarding the changed point and not recording; if li is smaller than or equal to lo, keeping the changed point, recording as one, and summarizing the relevant parameters of the point to obtain a data set Vi=(Xic,Yic,ai,bi,6i,ni,si), wherein the circle center of the candidate ellipse is (Xic, Yic), A semi-major axis is a, a semi-minor axis is bi, a rotation angle is 8;, sirepresents a serial number of each segment, and ni is the number of contour segments; and repeating the above comparison process untit all points on the candidate ellipse contour are compared, summing up all the kept data set of the M points to obtain
V={Vi=(xic,Yic,ai,bi,0i,ni,si)[i=1,2,...,M }, wherein the one with the most votes is determined as the candidate circle.
7. The clustering algorithm-based multi-parameter cumulative calculation method for lower limb vascular calcification indexes according to claim 6, wherein in the step 4.3.1, using a snake model to give a 2D parameter closed curve near the region of interest, and by minimizing the energy functional, deforming the closed curve in the image and continuously approximating the target contour, receiving the final evolution results as the target contour, and expressing a contour curve energy function as follows: Ef Eu os = Ep 5) + E OD 0 0 wherein, Esnake (V(S)) is curve energy, v(s) is a parameter equation of snake contour, Eint is internal energy of the curve, which determines the smoothness and continuity of the curve; Eex is energy given to the curve by the outside, which makes the curve move towards a characteristic direction of the target, and s is an independent variable describing the boundary; and in the step 4.3.2, using the least-square circle fitting method to re-fit the circle, and getting the center of the circle through a weighting function of the coordinates of the edge points on all the circles, that is, the centers (X, Y) 1 < | € X=—}x Y==>y of the calcified spots, wherein N°5 , NZ and the diameters
2 + 2 3 D==2 (x X) +0, =") N°5 of the calcified spots are calculated, xi and yi respectively represent the coordinates of certain point on the contours of the calcified spots, and N is the number of points on the contours of the calcified spots; and finally getting the area of the calcified spots.
8. The clustering algorithm-based multi-parameter cumulative calculation method for lower limb vascular calcification indexes according to claim 1, wherein the cumulative correction coefficient k is equal to kn*kw*Ko*kp, kn is a cumulative score of stenosis; kw is a cumulative score of wall shear stress w of the lower extremity arteries; ko is a cumulative score of an oscillating shear index o of the lower extremity arteries; kp is a cumulative score of wall shear stress p of the lower extremity arteries; the cumulative calculation standard for lower extremity arterial stenosis is as follows: according to the lumen area of the lower extremity arteries, it is divided into four grades: |. under lumen diameter reduction of 1%-25%, cumulative stenosis score kn is 1 point; Il. under lumen diameter reduction of 25%-50%, stenosis cumulative score kn is 2 points; Ill. under lumen diameter reduction 51%-75%, stenosis cumulative score kn is 3 points: IV. under lumen diameter reduction 76%-100%, stenosis cumulative score kn is 4 points, wherein the method of collecting stenosis data is as follows: collecting the maximum value of the vascular stenosis hmax and the average value h of the stenosis, in the segment with calcified plaques, of the lower extremity artery, and determining the value h of vascular stenosis of this segment according to h= ah, +bh , wherein a and b are constant coefficients; the cumulative calculation standard for wall shear stress w of lower limb arteries is as follows: wall shear stress in an arterial system is generally (10-70) dynes/cm?, and when wall shear stress w is (0-4) dynes/cm?, the cumulative score kw is 3 points, when the wall shear stress w is (5-10) dynes/cm?, the cumulative score kw is 2 points, and when the wall shear stress w is (11-70) dynes/cm?, the cumulative score kw is 1 point; the accumulative calculation standard for the oscillating shear index o of the lower limb arteries is as follows: the normal range of the index is 0-0.5, when the oscillating shear index is lower than 0.2, the cumulative score ko is 3 points, when the oscillating shear index is between 0.2 and 0.3, the cumulative score ko is 2 points, and when the oscillating shear index is between 0.3 and
0.5, the cumulative score ko is 1 point; the accumulative calculation standard for wall shear stress p of lower limb arteries is as follows: a normal range of indicators is systolic blood pressure of 90-140mmHg, diastolic blood pressure of 60-90mmHg; when the systolic blood pressure is lower than 90mmHg and the diastolic blood pressure is lower than 60mmHg, the cumulative score kp is 3 points; when the systolic blood pressure is between 90-140mmHg and the diastolic blood pressure is between 60-90mmHg, the cumulative score kp is 2 points; and when the systolic blood pressure is higher than 140mmHg and the diastolic blood pressure is higher than 90mmHg, the cumulative score kp is 1 point.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
LU102573A LU102573B1 (en) | 2021-02-24 | 2021-02-24 | Clustering Algorithm-based Multi-parameter Cumulative Calculation Method for Lower Limb Vascular Calcification Indexes |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
LU102573A LU102573B1 (en) | 2021-02-24 | 2021-02-24 | Clustering Algorithm-based Multi-parameter Cumulative Calculation Method for Lower Limb Vascular Calcification Indexes |
Publications (1)
Publication Number | Publication Date |
---|---|
LU102573B1 true LU102573B1 (en) | 2021-08-24 |
Family
ID=77411441
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
LU102573A LU102573B1 (en) | 2021-02-24 | 2021-02-24 | Clustering Algorithm-based Multi-parameter Cumulative Calculation Method for Lower Limb Vascular Calcification Indexes |
Country Status (1)
Country | Link |
---|---|
LU (1) | LU102573B1 (en) |
-
2021
- 2021-02-24 LU LU102573A patent/LU102573B1/en active IP Right Grant
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11557072B2 (en) | Clustering algorithm-based multi-parameter cumulative calculation method for lower limb vascular calcification indexes | |
CN110490040B (en) | Method for identifying local vascular stenosis degree in DSA coronary artery image | |
US20130016092A1 (en) | Method and system of generating a 3d visualization from 2d images | |
CN104299242B (en) | Fluoroscopic visualization eye fundus image extracting method based on NGC ACM | |
CN108618749B (en) | Retina blood vessel three-dimensional reconstruction method based on portable digital fundus camera | |
CN107909590B (en) | IVUS image outer membrane edge segmentation method based on Snake improved algorithm | |
CN109389568B (en) | Method for automatically measuring skin thickness in skin optical coherence tomography image | |
CN110448319A (en) | Based on radiography image and blood flow velocity calculation method coronarius | |
Liu et al. | Cardiac magnetic resonance image segmentation based on convolutional neural network | |
WO2023151280A1 (en) | Three-dimensional fusion method and fusion system for dual-mode coronary artery blood vessel images | |
CN113470137A (en) | IVOCT image guide wire artifact removing method based on gray-scale weighting | |
CN112562058B (en) | Method for quickly establishing intracranial vascular simulation three-dimensional model based on transfer learning | |
CN112950737A (en) | Fundus fluorescence radiography image generation method based on deep learning | |
CN110353639B (en) | Blood supply area quantification method and system based on blood vessel enhanced radiography | |
CN111445473A (en) | Vascular membrane accurate segmentation method and system based on intravascular ultrasound image sequence multi-angle reconstruction | |
Jainish et al. | A novel automatic retinal vessel extraction using maximum entropy based EM algorithm | |
LU102573B1 (en) | Clustering Algorithm-based Multi-parameter Cumulative Calculation Method for Lower Limb Vascular Calcification Indexes | |
CN113537300A (en) | Deep learning staging method based on diabetes foot wound ischemia or infection image | |
Hennersperger et al. | Vascular 3D+ T freehand ultrasound using correlation of doppler and pulse-oximetry data | |
CN111292285B (en) | Automatic screening method for diabetes mellitus based on naive Bayes and support vector machine | |
CN109785296B (en) | CTA image-based three-dimensional spherical index determination method | |
CN108154512A (en) | It is a kind of based on the multiple retinal images blood vessel segmentation system for going trend analysis | |
CN115147273A (en) | Domain-adaptation-based generation method and system for ophthalmologic OCT (optical coherence tomography) postoperative profile | |
Mi et al. | Detecting carotid intima-media from small-sample ultrasound images | |
Archana et al. | Detection of abnormal blood vessels in diabetic retinopathy based on brightness variations in SDOCT retinal images |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
FG | Patent granted |
Effective date: 20210824 |