WO2021238739A1 - 基于聚类算法的下肢血管钙化指数多参数累积计算方法 - Google Patents

基于聚类算法的下肢血管钙化指数多参数累积计算方法 Download PDF

Info

Publication number
WO2021238739A1
WO2021238739A1 PCT/CN2021/094611 CN2021094611W WO2021238739A1 WO 2021238739 A1 WO2021238739 A1 WO 2021238739A1 CN 2021094611 W CN2021094611 W CN 2021094611W WO 2021238739 A1 WO2021238739 A1 WO 2021238739A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
cumulative
points
lower extremity
calcified
Prior art date
Application number
PCT/CN2021/094611
Other languages
English (en)
French (fr)
Inventor
王中群
耿跃
李丽华
邵晨
孙振
臧光耀
张莉莉
Original Assignee
江苏大学附属医院
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 江苏大学附属医院 filed Critical 江苏大学附属医院
Priority to US17/432,094 priority Critical patent/US11557072B2/en
Publication of WO2021238739A1 publication Critical patent/WO2021238739A1/zh

Links

Images

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
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
    • A61B6/50—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications
    • A61B6/504—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications for diagnosis of blood vessels, e.g. by angiography
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
    • A61B6/02—Arrangements for diagnosis sequentially in different planes; Stereoscopic radiation diagnosis
    • A61B6/03—Computed tomography [CT]
    • A61B6/032—Transmission computed tomography [CT]
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
    • A61B6/52—Devices using data or image processing specially adapted for radiation diagnosis
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
    • A61B6/52—Devices using data or image processing specially adapted for radiation diagnosis
    • A61B6/5211—Devices using data or image processing specially adapted for radiation diagnosis involving processing of medical diagnostic data
    • A61B6/5217—Devices using data or image processing specially adapted for radiation diagnosis involving processing of medical diagnostic data extracting a diagnostic or physiological parameter from medical diagnostic data
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
    • A61B6/52—Devices using data or image processing specially adapted for radiation diagnosis
    • A61B6/5258—Devices using data or image processing specially adapted for radiation diagnosis involving detection or reduction of artifacts or noise
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T11/00—2D [Two Dimensional] image generation
    • G06T11/003—Reconstruction from projections, e.g. tomography
    • G06T11/008—Specific post-processing after tomographic reconstruction, e.g. voxelisation, metal artifact correction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/70—Denoising; Smoothing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/149—Segmentation; Edge detection involving deformable models, e.g. active contour models
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/36—Applying a local operator, i.e. means to operate on image points situated in the vicinity of a given point; Non-linear local filtering operations, e.g. median filtering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
    • G06V10/469—Contour-based spatial representations, e.g. vector-coding
    • G06V10/473—Contour-based spatial representations, e.g. vector-coding using gradient analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
    • G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
    • G06V10/76—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries based on eigen-space representations, e.g. from pose or different illumination conditions; Shape manifolds
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2211/00—Image generation
    • G06T2211/40—Computed tomography
    • G06T2211/424—Iterative
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/03—Recognition of patterns in medical or anatomical images

Definitions

  • the invention belongs to the technical field of medical image processing, and specifically relates to a multi-parameter cumulative calculation method for lower limb vascular calcification index based on a clustering algorithm.
  • Diabetic lower extremity arteriosclerosis obliterans is a more serious peripheral arterial vascular disease in which diabetes has developed to a certain degree. As shown in Figure 5, diabetes, hypertension, lipid metabolism disorders, and smoking are important risk factors for its onset. Due to the long-term stimulation of hyperglycemia in diabetic patients, the degree of peripheral arterial disease is often more serious than that of patients with simple hypertension, and patients with lower extremity arterial disease are more likely to cause higher disability and mortality. Mild cases can cause numbness and chills in the lower limbs, and moderate to severe cases can cause intermittent resting pain, and eventually develop into necrosis, leading to amputation and disability.
  • diabetic lower extremity arteriosclerosis obliterans tends to have more extensive and severe disease than those with lower extremity arteriosclerosis obliterans caused by other reasons, such as more plaques and more serious small vessel disease below the knee. Because of its complexity, it often brings certain difficulties to clinical treatment.
  • the examination techniques for lower extremity arteries include ankle-brachial index measurement, arteriography, color Doppler, CT angiography, magnetic resonance angiography and other methods.
  • the above-mentioned inspection techniques also have the following shortcomings.
  • the ankle-brachial index measurement cannot judge the degree and nature of vascular stenosis, and there is a false-negative rate when the calcification is more serious.
  • Arteriography is an invasive examination with high cost and many complications.
  • the results of color Doppler examination are greatly affected by factors such as the proficiency of the operator, and the display of deep blood vessels and adjacent bone blood vessels is poor.
  • the present invention proposes a multi-parameter accumulation calculation method for the lower extremity vascular calcification index based on a clustering algorithm.
  • the calcification area combined with the corresponding cumulative correction coefficient obtained from the vascular fluid mechanics study, gives a quantitative value of the vascular calcification index of the lower extremity.
  • the value of the vascular calcification index of the lower extremity can represent the degree of calcification of the lower extremity blood vessels to a large extent. Diabetic foot faces the diagnostic risk of amputation to provide data basis.
  • Step 1 Acquire CT images of blood vessels of the lower extremities to be analyzed
  • Step 2 Use linear iterative clustering algorithm to evenly segment the calcified spots in the CT image in each super pixel area;
  • Step 3 After completing the super pixel segmentation, extract the CT brightness feature value of the super pixel area where the calcified spot is located through the Lab color space.
  • Step 4 Perform edge detection and contour extraction on the calcified spots in the CT image, use segmented ellipses to fit the calcified spots in the processed image and perform optimization processing to obtain the radius of the calcified spots, and then calculate the area of the calcified spots;
  • step 2 the method for performing super pixel segmentation on the CT image in step 2 is:
  • Step 2.1 Perform uniform super-pixel segmentation on the collected original lower extremity blood vessel CT image.
  • X pixels in the original lower extremity blood vessel CT image, and the original lower extremity blood vessel CT image is divided into K, then each super pixel has Pixels;
  • Step 2.2 preset the interval between the initial cluster center C and the initial cluster center C;
  • Step 2.3 search for pixels close to C in the field of cluster center C based on Euclidean distance, and classify these pixels into one category;
  • Step 2.4 Calculate the average eigenvector value of all pixels in the K superpixel area, perform the next clustering based on the average eigenvector value, iteratively update the cluster center, and iterate again until the end of the iteration;
  • Step 2.5 segment the super pixel after the iteration to obtain the super pixel area.
  • the method for extracting the brightness feature value of the super pixel area where the calcified spot is located in the step 3 is:
  • Step 3.1 extract the brightness channel L in Lab to characterize the brightness characteristics of CT images of blood vessels in the lower extremities Then obtain the brightness image L 0 of the lower extremity blood vessel CT image, Y is the intermediate variable, Y 0 is the gray value of white defined by the CIE standard, and f is the correction function;
  • Step 3.2 Extract the brightness feature value of the super pixel area where the calcified spot is located based on the brightness map L 0.
  • step 3.2 the process of extracting the brightness feature value in step 3.2 is:
  • Step 3.2.2 extract the maximum brightness pixel A(x,y) in the super pixel area after Gaussian filtering, and obtain the sum of the gray values of all pixels in the super pixel area
  • (x, y) are the pixel coordinates
  • y ⁇ (y 1 , y 2 ) x ⁇ (x 1 , x 2 )
  • y 1 , y 2 are the coordinate values of the y-axis direction in the super pixel area, respectively
  • x 1 and x 2 are the coordinate values in the x- axis direction in the super pixel area
  • f(x,y) is the pixel value at (x,y)
  • P y (x) is the column with the abscissa at x Vector pixel gray value cumulative sum, and then pass Obtain the cumulative sum of the gray values of all pixels in the entire super pixel area;
  • Step 3.2.3 the cumulative sum of the gray values of all pixels in the entire super pixel area is the CT brightness value.
  • Step 4.1.1 use Gaussian filter to preprocess the image
  • Step 4.1.2 use the sobel operator to calculate the gradient size and direction of each pixel in the filtered image
  • Step 4.1.3 select edge points based on the comparison of the gradient intensity of the pixels.
  • the gradient intensity of a pixel is greater than the other two pixels along the positive and negative gradient direction, it will be retained as an edge point, otherwise the pixel will be suppressed;
  • Step 4.1.4 compare the edge points obtained in the previous step with the upper threshold, and then filter the edge points; if the upper threshold is less than the edge point, keep the point and set the changed point as the first edge point; then search Whether the neighboring points of this point are less than the upper threshold, repeat this process, and connect all the points greater than the upper threshold.
  • Step 4.2.1 randomly divide the obtained contour into n segments
  • Step 4.2.2 arbitrarily select 12 non-repeated points in each segment of the contour, and use the least squares method to fit n candidate ellipses;
  • the active contour model is used to optimize the contour of the ellipse to obtain the area of the calcified spots
  • Step 4.3.1 using the snake model to give a 2D parameter closed curve near the region of interest, by minimizing the energy functional, this closed curve is deformed in the image and continuously approximates the target contour, and the final evolution result is received as The target contour, the contour curve energy function is expressed as:
  • E snake (v(s)) is the curve energy
  • v(s) is the parameter equation of the snake profile
  • E int is the internal energy of the curve, which determines the smoothness and continuity of the curve
  • E ext is the curve given by the outside world Energy, which makes the curve move towards the characteristic direction of the target
  • s is the independent variable describing the boundary
  • Step 4.3.2 use the least squares circle fitting method to re-fit the circle, and get the center of the circle through the weighting function of the edge point coordinates on all circles, that is, the center (X, Y) of the calcified spot, where, And then calculate the diameter of the calcified spots Among them, x i and y i respectively represent the coordinates of a certain point on the contour of the calcification spot, and N is the number of points on the contour of the calcification spot; finally, the area of the calcification spot is obtained.
  • the cumulative correction coefficient k k h *k w *k o *k p , k h is the cumulative stenosis score; k w is the cumulative score of the wall shear stress w of the lower extremity arteries; k o is the lower extremity arterial vascular oscillating shear index o Cumulative score; k p is the cumulative score of lower extremity arterial wall pressure p;
  • Calculating a standard cumulative degree lower extremity artery stenosis The lower extremity arterial vessel lumen area, divided into four levels: I, luminal diameter narrowing 1% to 25%, the cumulative score of stenosis Comments h k 1 min, II, Lumen diameter reduction 25%-50%, stenosis cumulative score k h score 2 points, III, lumen diameter reduction 51%-75%, stenosis cumulative score k h score 3 points, IV, lumen diameter reduction 76% -100%, the cumulative stenosis score k h is rated 4 points; the method of collecting stenosis data is: collect the maximum value of the vascular stenosis h max and the average value of the stenosis h of the lower extremity arteries in the calcified plaque segment, according to Determine the degree of vascular stenosis in this segment Where a and b are constant coefficients;
  • the cumulative calculation standard for wall shear stress w of lower extremity arteries is: the wall shear stress in the arterial system is generally (10 ⁇ 70) dynes/cm 2 , and the cumulative score k w when the wall shear stress w is (0 ⁇ 4) dynes/cm 2 When the wall shear stress w is (5 ⁇ 10) dynes/cm 2 , the cumulative score k w is 2 points, and when the wall shear stress w is (11 ⁇ 70) dynes/cm 2 , the cumulative score k w is 1 point;
  • the standard for calculating the oscillating shear index of lower extremity arteries is: the normal range of the index: 0 ⁇ 0.5, when the oscillating shear index is lower than 0.2, the cumulative score k o is 3 points, and when the oscillating shear index is between 0.2 and 0.3, the cumulative score k o is 2 points, and the cumulative score k o is 1 point when the shock shear index is between 0.3 and 0.5;
  • Lower extremity arterial wall pressure p is calculated cumulative criteria: the normal range indicators: systolic blood pressure: 90 ⁇ 140mmHg, diastolic blood pressure: 60 ⁇ 90mmHg; systolic blood pressure less than 90mmHg, diastolic blood pressure less than 60mmHg cumulative score K p is 3 minutes, systolic blood pressure between 90 ⁇ 140mmHg, diastolic blood pressure between 60 ⁇ 90mmHg cumulative score k p is 2 minutes, systolic blood pressure greater than 140mmHg, diastolic blood pressure greater than 90mmHg cumulative score k p is 1 min.
  • the multi-parameter accumulation calculation method of lower extremity blood vessel calcification index proposed by the present invention is based on the lower extremity blood vessel CT image processing process by first performing super pixel segmentation on the lower extremity blood vessel CT image.
  • the pixels can be aggregated in At the same time, multiple sub-region blocks with regular shapes and consistent local structures are formed, which realizes the overall expression of image local features and structural information, avoids excessive data, increases processing speed, and uses superpixels to reduce data dimensions.
  • 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 retain effective information to the greatest extent and reduce noise.
  • the present invention uses the Lab color space to extract the brightness feature value of the super pixel area where the calcified spots are located.
  • the original image L 0 is layer-by-layer low-pass filter and sub-sampling operation to obtain the original image L 0 spatial scale transformation brightness intensity Figure, you can enhance the edge of the salient area of the image.
  • the sum of the gray values of all pixels in the super pixel area is calculated according to the pixel with the maximum brightness in the super pixel area, and the CT brightness value can be obtained from the Lab color space by using the pixel gray value integral.
  • the Canny operator is used to perform edge detection and contour extraction on the calcified spots in the image, and the Gaussian filter is used to preprocess the image to reduce the effect of noise; use non-maximum suppression to make the edge There is an accurate response, the edge is detected at the correct position, and the accuracy of the target is improved.
  • Use hysteresis threshold processing detection to connect edge points, remove false edges, and improve edge positioning accuracy.
  • the present invention uses the active contour model to optimize the elliptical 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 the final evolution result is received as the target contour, and passes through the snake model The contour after curve evolution and refitting is closer to the real contour.
  • the multi-parameter cumulative calculation method of the lower extremity blood vessel calcification index of the present invention based on the clustering algorithm is a new way of calculating the CT brightness characteristic value and calcification area obtained by the CT image processing of the lower extremity blood vessel, combined with the corresponding cumulative correction obtained by the vascular fluid mechanics study
  • the coefficient gives a quantitative value of the vascular calcification index of the lower extremity.
  • the value of the vascular calcification index of the lower extremity can represent the degree of calcification of the lower extremity blood vessels to a large extent, providing data basis for the subsequent diagnosis of diabetic foot facing amputation risk, effectively reducing empirical judgments The error caused.
  • the degree of calcification is obtained by processing the CT image, which avoids the influence of human judgment factors and improves the accuracy.
  • Figure 1 is a process flow chart of the method of the present invention
  • Figure 2 is a flowchart of dividing the super pixel area
  • Figure 3 is a CT image of the lower limbs of a diabetic patient
  • Figure 4 is a schematic diagram of arterial vascular media calcification and intimal calcification
  • Figure 5 is a schematic diagram of the progression of atherosclerosis.
  • Step 1 Use CT and other testing instruments to collect CT images of blood vessels in the lower extremities of diabetic patients, and obtain CT images of blood vessels in the lower extremities that need to be analyzed;
  • Step 2 This application uses a linear iterative clustering algorithm (SLIC) to uniformly segment the calcified spots in the CT image into each superpixel area; the process is as follows:
  • SLIC linear iterative clustering algorithm
  • Step 2.1 suppose that the original CT image of lower extremity blood vessels has X pixels, set the number of initial superpixels to K, and divide the original CT image uniformly. At this time, each superpixel area that is segmented has Pixels;
  • Step 2.2 preset the interval between the initial cluster center C and the initial cluster center C;
  • Step 2.3 search for pixels close to C in the field of cluster center C based on Euclidean distance, and classify these pixels into one category, until all pixels in the field of each cluster center C are divided;
  • Step 2.4 Calculate the average eigenvector value of all pixels in the K superpixel area, perform the next clustering based on the average eigenvector value, iteratively update the cluster center, and iterate again until the end of the iteration.
  • Step 2.5 segment the super pixel after the iteration to obtain the super pixel area.
  • Step 3 After completing the super pixel segmentation, extract the brightness feature value of the super pixel area where the calcified spot is located through the Lab color space.
  • the specific process is:
  • Step 3.1 extract the brightness channel L in Lab to characterize the brightness characteristics of CT images of blood vessels in the lower extremities Then obtain the brightness image L 0 of the lower extremity blood vessel CT image, Y is the intermediate variable, Y 0 is the gray value of white defined by the CIE standard, and f is the correction function;
  • Step 3.2 extract the brightness feature value of the super pixel area where the calcified spot is located based on the L channel brightness map L 0;
  • Step 3.2.2 extract the maximum brightness pixel A(x, y) in all super pixel areas in the image L 1 after brightness enhancement, and obtain the sum of the gray values of all pixels in the super pixel area; namely: Among them, (x, y) are the pixel coordinates, y ⁇ (y 1 , y 2 ), x ⁇ (x 1 , x 2 ), y 1 , y 2 are the coordinate values of the y-axis direction in the super pixel area, respectively, x 1 and x 2 are the coordinate values in the x- axis direction in the super pixel area, f(x,y) is the pixel value at (x,y), and P y (x) is the column with the abscissa at x Vector pixel gray value cumulative sum, and then pass Obtain the cumulative sum of the gray values of all pixels in the entire super pixel area;
  • Step 3.2.3 the cumulative sum of the gray values of all pixels in the entire super pixel area is the CT brightness value.
  • Step 4 Extract the calcification area in the CT image. It can be seen from the CT image that the calcification spots distributed in the patient's lower limbs appear as spots with matte periphery. Therefore, the following processing is required to extract the calcification area from the image:
  • Step 4.1 the process method of processing calcified spots using edge detection and contour extraction is:
  • Step 4.1.1 first, use Gaussian filter to preprocess the image, which can smooth the image and filter the noise at the same time
  • Step 4.1.2 use the sobel operator to calculate the gradient size and direction of each pixel in the filtered image
  • Step 4.1.3 select the edge points based on the gradient intensity comparison of the pixels.
  • the gradient intensity of a pixel is greater than the other two pixels along the positive and negative gradient direction, it is retained as an edge point, otherwise the pixel is suppressed, that is, the The point suppression is 0; through the above selection process, the edge point can produce an accurate response, which improves the accuracy of edge point extraction.
  • Step 4.1.4 set the upper threshold, compare the edge points obtained in the previous step with the upper threshold, and then filter the edge points; if the upper threshold is less than the edge point, keep this point and set the changed point as the first Edge point;; Then find whether the neighboring point of the point is greater than the upper threshold, repeat this process, and connect all the points greater than the upper threshold, through the hysteresis threshold processing can remove false edges, so that the edge positioning accuracy is improved.
  • Step 4.2 the method of fitting the segmented ellipse is:
  • Step 4.2.1 randomly divide the contour obtained in the previous step into n segments, n ⁇ [8, 12], where n is an even number in the range;
  • Step 4.2.2 randomly select 12 non-repeated points in each segment of the contour, and use the least squares method to fit n candidate ellipses;
  • Step 4.3 Use the active contour model to optimize the ellipse contour, and finally obtain the area of calcified spots;
  • Step 4.3.1 use the snake model to give a 2D parameter closed curve near the region of interest.
  • the closed curve is deformed in the image and continuously approximates the target contour, and the final evolution result is received Is the target contour, the contour curve energy function is expressed as:
  • E snake (v(s)) is the curve energy
  • v(s) is the parameter equation of the snake profile
  • E int is the internal energy of the curve, which determines the smoothness and continuity of the curve
  • E ext is the curve given by the outside world Energy, which makes the curve move toward the characteristic direction of the target
  • s is the independent variable describing the boundary.
  • Step 4.3.2 use the least squares circle fitting method to re-fit the circle, and get the center of the circle through the weighting function of the edge point coordinates on all circles, that is, the center (X, Y) of the calcified spot, where, And then calculate the diameter of the calcified spots Among them, x i and y i respectively represent the coordinates of a certain point on the contour of the calcification spot, and N is the number of points on the contour of the calcification spot; finally, the area of the calcification spot is obtained.
  • the index is not only related to the area S of the calcified spots in the CT image, the brightness of the calcified spots, and the degree of stenosis, but also closely related to the wall shear stress w, the oscillating shear index o, and the wall pressure p.
  • the 4cm length of the area where the calcified spots is intercepted uses the cumulative correction coefficient k to correct the degree of calcification, because the types of calcification in the lower extremity arteries include intimal calcification and media calcification, and according to the calcification spots in the CT image, it is impossible to directly distinguish the intima Calcification or media calcification is shown in Figures 4 and 5, but intimal calcification can easily cause stenosis of the arterial lumen.
  • the cumulative calculation of the area S of the lower extremity arterial calcified plaques can be directly obtained by fitting the calcified spots in the processed image using a segmented ellipse and performing optimization processing to obtain the radius of the calcified spots, and then calculating the area of the calcified spots; or Directly given by the scoring standard: the size of the calcified plaque area on the anterior and posterior walls of the lower extremity arteries is divided into four grades: I, no calcification, score 0; II, calcification range less than 1/3 of the arterial wall length, Score 1 point; III. Calcification range involving 1/3 to 2/3 of the arterial wall length, score 2 points; IV.
  • the blood vessels of the lower extremities are divided into the upper and lower sections in the middle of the calf, including: proximal posterior tibial artery, distal posterior tibial artery, proximal anterior tibial artery, distal anterior tibial artery, proximal peroneal artery, and distal peroneal artery. end.
  • the brightness value ⁇ of the super pixel area where the calcified spot is located can be directly obtained by extracting the CT brightness feature value of the super pixel area where the calcified spot is located through the Lab color space in the clustering algorithm.
  • the cumulative calculation standard for wall shear stress w of lower extremity arteries is: the wall shear stress in the arterial system is generally (10 ⁇ 70) dynes/cm 2 , and the cumulative score k w when the wall shear stress w is (0 ⁇ 4) dynes/cm 2 When the wall shear stress w is (5 ⁇ 10) dynes/cm 2 , the cumulative score k w is 2 points, and when the wall shear stress w is (11 ⁇ 70) dynes/cm 2 , the cumulative score k w is 1 point.
  • the standard for calculating the oscillating shear index of lower extremity arteries is: the normal range of the index: 0 ⁇ 0.5, when the oscillating shear index is lower than 0.2, the cumulative score k o is 3 points, and when the oscillating shear index is between 0.2 and 0.3, the cumulative score k o is 2 points, and the cumulative score k o is 1 point when the shock shear index is between 0.3 and 0.5.
  • Lower extremity arterial wall pressure p is calculated cumulative criteria: the normal range indicators: systolic blood pressure: 90 ⁇ 140mmHg, diastolic blood pressure: 60 ⁇ 90mmHg; systolic blood pressure less than 90mmHg, diastolic blood pressure less than 60mmHg cumulative score K p is 3 minutes, systolic blood pressure between 90 ⁇ 140mmHg, diastolic blood pressure between 60 ⁇ 90mmHg cumulative score k p is 2 minutes, systolic blood pressure greater than 140mmHg, diastolic blood pressure greater than 90mmHg cumulative score k p is 1 min.
  • the multi-parameter cumulative calculation method of the lower extremity vascular calcification index based on the clustering algorithm of the present invention does not directly aim to obtain the diagnosis result or health status of diabetic foot disease, but is only a method of obtaining information as an intermediate result from a living human body, which is diabetes
  • the diagnosis of foot diseases provides intermediate data support, which does not belong to the category of disease diagnosis.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Health & Medical Sciences (AREA)
  • Multimedia (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Molecular Biology (AREA)
  • Biomedical Technology (AREA)
  • High Energy & Nuclear Physics (AREA)
  • Animal Behavior & Ethology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Public Health (AREA)
  • Optics & Photonics (AREA)
  • Veterinary Medicine (AREA)
  • Biophysics (AREA)
  • Surgery (AREA)
  • Pathology (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Vascular Medicine (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Dentistry (AREA)
  • Software Systems (AREA)
  • Databases & Information Systems (AREA)
  • Computing Systems (AREA)
  • Pulmonology (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • General Engineering & Computer Science (AREA)
  • Nonlinear Science (AREA)
  • Physiology (AREA)
  • Quality & Reliability (AREA)

Abstract

基于聚类算法的下肢血管钙化指数多参数累积计算方法，对首先对CT图像进行超像素分割，将CT图像中钙化斑点分割在各个超像素区域内；在完成超像素分割后，使用Lab颜色空间提取钙化斑点所在超像素区域的亮度特征值，并对图像中钙化斑点进行边缘检测和轮廓提取，经过边缘检测和轮廓提取后，对图像中的钙化斑点使用分段椭圆进行拟合及优化椭圆轮廓后对钙化斑点的面积进行提取；基于亮度特征值、钙化斑点的面积和累积校正系数等多参数从获得CT图像中钙化程度的判断值，直接对下肢血管CT图像进行处理有效降低经验判断的误差。

Description

基于聚类算法的下肢血管钙化指数多参数累积计算方法 技术领域
本发明属于医学图像处理技术领域，具体涉及基于聚类算法的下肢血管钙化指数多参数累积计算方法。
背景技术
糖尿病下肢动脉硬化闭塞症是糖尿病发展到一定程度的一种较为严重的外周动脉血管性疾病如图5所示，糖尿病、高血压、脂代谢紊乱、吸烟等是其发病的重要危险因素。由于糖尿病患者长期处在高血糖因素的刺激下，其外周动脉病变程度常较单纯高血压病人更为严重，其下肢动脉病变患者更容易引起较高的致残率及死亡率。轻者可引起下肢麻木、发凉，中重度则可引起间歇性静息痛，最后发展为坏死从而导致截肢致残。由于糖尿病病情的复杂性和病程的长期性，糖尿病下肢动脉硬化闭塞症往往较其他原因引起的下肢动脉硬化闭塞患者病变较为广泛且严重，如斑块较多，且膝以下小血管病变更为严重，由于其复杂性，往往给临床治疗工作带来了一定的困难。
目前，对下肢动脉的检查技术包括踝肱指数测定、动脉造影、彩色多普勒、CT血管成像、磁共振血管造影等手段。然后上述检查技术也存在着以下不足，踝肱指数测定无法对对血管狭窄程度及性质做出判断，当钙化比较严重时存在假阴性率。动脉造影是一项由创性检查且成本高，并发症多。彩色多普勒的检查结果受操作者熟练程度等因素影响较大，且对深部血管及临近骨头血管显示较差。磁共振血管造影的应用虽然越来越广泛，但是磁共振血管造影空间分辨率低，对细小的血管诊断仍存在较大偏差，不能完全达到临床需要。CT图像技术在血管成像的应用越来越被广泛应用，但是临床医师根据成像结果对患者的病情进行评估，大部分是依靠医师的个人经验进行的判断，所以判断结果受到很大程度人为因素的影响，因此针对CT图像需要设计一种更有效的方法，通过对CT图像的处理，输出较为准确下肢血管钙化程度，有效避免人为因素影响判断结果。
除此之外，和人体其他动脉钙化斑块分布不同的是，下肢动脉血管分支分化更加细化和繁杂，由于糖尿病足面临截肢的诊断准确性往往受到采集下肢血管内的钙化斑块的钙化程度的影响，如何正确获取下肢血管钙化指数将是目前研究的重点和难点。
发明内容
为了解决现有技术中下肢血管钙化指数缺乏有效的计算手段的问题，本发明提出了基 于聚类算法的下肢血管钙化指数多参数累积计算方法，通过对下肢血管的CT图像处理得到的CT亮度特征值、钙化面积，结合血管流体力学研究获得的相应的累积校正系数，给出了下肢血管钙化指数的量化数值，该下肢血管钙化指数的数值能够很大程度的表征下肢血管的钙化程度，为后续的糖尿病足面临截肢的诊断风险提供数据依据。
本发明所采用的技术方案如下：
步骤1，采集待分析的下肢血管CT图像；
步骤2，使用线性迭代聚类算法将CT图像中钙化斑点被均匀分割在每一个超像素区域内；
步骤3，完成超像素分割后，通过Lab颜色空间提取钙化斑点所在超像素区域的CT亮度特征值，
步骤4，对CT图像中的钙化斑点进行边缘检测和轮廓提取，利用分段椭圆对处理后的图像中钙化斑点进行拟合并进行优化处理得到钙化斑点的半径，进而计算钙化斑点的面积；
步骤5，根据Cal＝k·ρ·S获得CT图像中钙化程度的判断值，其中，ρ为CT亮度特征值，S为钙化斑点的面积，k为累积校正系数。
进一步，所述步骤2将CT图像进行超像素分割的方法为：
步骤2.1，对所采集的原始下肢血管CT图像进行均匀的超像素分割，设原始下肢血管CT图像中的像素共有X个，原始下肢血管CT图像被分割成K个，则每个超像素都有
Figure PCTCN2021094611-appb-000001
个像素；
步骤2.2，预设初始的聚类中心C和初始聚类中心C之间的间隔；
步骤2.3，基于欧几里距离搜寻聚类中心C领域内靠近C的像素点，并将这些像素点划为一类；
步骤2.4，计算K个超像素区域内所有像元的平均特征向量值，根据平均特征向量值进行下一次聚类，迭代更新聚类中心，再次迭代直到迭代结束；
步骤2.5，对完成迭代后的超像素进行分割，获得超像素区域。
进一步，所述步骤3中提取钙化斑点所在超像素区域的亮度特征值的方法为：
步骤3.1，提取出Lab中的亮度通道L，表征下肢血管CT图像亮度特征
Figure PCTCN2021094611-appb-000002
进而获得下肢血管CT图像的亮度图像L 0，Y为的中间变量，Y 0为是CIE标准定义白色的灰度值，f为矫正函数；
步骤3.2，基于亮度图L 0提取钙化斑点所在超像素区域的亮度特征值。
进一步，步骤3.2提取亮度特征值的过程为：
步骤3.2.1，使用高斯滤波1/2下采样对L 0进行处理后得到图像L 1，L 1＝subsample(lpfilter(L 0))，其中，subsample()为下采样函数，lpfilter()为频域滤波函数；
步骤3.2.2，提取经高斯滤波后的超像素区域内亮度最大像素点A(x,y)，并求得该超像素区域内所有像素灰度值的和
Figure PCTCN2021094611-appb-000003
其中，(x,y)为像素坐标，y∈(y 1,y 2)，x∈(x 1,x 2)，y 1、y 2分别是该超像素区域内y轴方向的坐标值，x 1、x 2分别是该超像素区域内 x轴方向的坐标值，f(x,y)为(x,y)处的像素值，P y(x)为在横坐标在 x处的列向量像素灰度值累加和，再通过
Figure PCTCN2021094611-appb-000004
得到整个超像素区域内所有像素灰度值累加和；
步骤3.2.3，整个超像素区域内所有像素灰度值的累加和即为CT亮度值。
进一步，利用边缘检测和轮廓提取对钙化斑点处理的过程法为：
步骤4.1.1，利用高斯滤波器对图像进行预处理；
步骤4.1.2，使用sobel算子计算滤波后图像中每个像素点的梯度大小和方向；
步骤4.1.3，基于像素的梯度强度对比选择边缘点，当某一像素的梯度强度大于沿正负梯度方向上的另外两个像素，则保留其为边缘点，反之则抑制该像素；
步骤4.1.4，将上一步得到的边缘点与上限阈值相比较，进而对边缘点进行筛选；若上限阈值小于边缘点，则保留该点并设定改点为第一个边缘点；接着查找该点的相邻点是否小于上限阈值，重复这一个过程，并将所有大于上限阈值的点进行连接。
进一步，所述分段椭圆进行拟合的方法为：
步骤4.2.1，将得到的轮廓随机分为n段；
步骤4.2.2，在每一段轮廓中任意选取12个不重复的点，利用最小二乘法拟合得到n个 候选椭圆；
步骤4.2.3，设置判断阈值l 0，将点(x i,y i)与候选椭圆轮廓之间的距离l i与判断阈值l 0进行比较，若l i>l 0，则丢弃改点，不记票数，若l i≤l 0，则保留改点，记1票，并将该点的相关参数汇总得到数据集V i＝(x ic,y ic,a i,b i,θ i,n i,s i)，其中，候选椭圆的圆心为(x ic,y ic)，半长轴为a i，半短轴为b i，旋转角为θ i，s i代表每个段的序列号，n i为轮廓的段数；重复上述比较过程直至所有候选椭圆轮廓上的点都被比较过，将所有被保留的M个点的数据集汇总得到V＝{V i＝(x ic,y ic,a i,b i,θ i,n i,s i)|i＝1,2,......,M}，且记票数最多的是拟合后的结果。
进一步，利用主动轮廓模型优化椭圆轮廓，进而得到钙化斑点的面积；
步骤4.3.1，利用snake模型在感兴趣区域附近给出一条2D参数闭合曲线，通过最小化能量泛函，让这条闭合曲线在图像中发生变形并不断逼近目标轮廓，最终演化结果被接收为目标轮廓，轮廓曲线能量函数表示为：
Figure PCTCN2021094611-appb-000005
其中，E snake(v(s))是曲线能量，v(s)是snake轮廓的参数方程，E int是曲线的内部能量，它决定曲线的光滑度和连续性；E ext是外界给曲线的能量，它使得曲线朝着目标的特征方向移动，s是描述边界的自变量；
步骤4.3.2，使用最小二乘圆拟合方法重新拟合圆，通过所有圆上的边缘点坐标的加权函数得到圆的中心，即得到钙化斑点的中心(X,Y)，其中，
Figure PCTCN2021094611-appb-000006
进而计算出钙化斑点的直径
Figure PCTCN2021094611-appb-000007
其中，x i和y i分别表示钙化斑点轮廓上某一点的坐标，N为钙化斑点轮廓上点的个数；最终的得出得到钙化斑点的面积。
进一步，
所述累积校正系数k＝k h*k w*k o*k p，k h为狭窄度累积评分；k w为下肢动脉血管壁面剪应力w累积评分；k o为下肢动脉血管震荡剪切指数o累积评分；k p为下肢动脉血管壁面压力p累积评分；
下肢动脉血管狭窄度的累积计算标准为：根据下肢动脉血管的管腔面积，分为四个等级：Ⅰ、管腔直径缩小1％-25％，狭窄度累积评分k h评1分，Ⅱ、管腔直径缩小25％-50％， 狭窄度累积评分k h评2分，Ⅲ、管腔直径缩小51％-75％，狭窄度累积评分k h评3分，Ⅳ、管腔直径缩小76％-100％，狭窄度累积评分k h评4分；狭窄度数据的采集方法为：采集下肢动脉在有钙化斑块段的血管狭窄度的最大值h max和狭窄度的平均值h，根据
Figure PCTCN2021094611-appb-000008
确定该段的血管狭窄度
Figure PCTCN2021094611-appb-000009
其中a、b为常系数；
下肢动脉血管壁面剪应力w累积计算标准为：动脉系统中的壁面剪应力一般为(10～70)dynes/cm 2，壁面剪应力w为(0～4)dynes/cm 2时累积评分k w为3分，壁面剪应力w为(5～10)dynes/cm 2时累积评分k w为2分，壁面剪应力w为(11～70)dynes/cm 2时累积评分k w为1分；
下肢动脉血管震荡剪切指数o累积计算标准为：指标正常范围：0～0.5，震荡剪切指数低于0.2时累积评分k o为3分，震荡剪切指数介于0.2～0.3时累积评分k o为2分，震荡剪切指数介于0.3～0.5时累积评分k o为1分；
下肢动脉血管壁面压力p累积计算标准为：指标正常范围：收缩压：90～140mmHg，舒张压：60～90mmHg；收缩压低于90mmHg、舒张压低于60mmHg累积评分k p为3分，收缩压介于90～140mmHg、舒张压介于60～90mmHg累积评分k p为2分，收缩压高于140mmHg、舒张压高于90mmHg累积评分k p为1分。
本发明的有益效果：
1、本发明所提出的基于聚类算法的下肢血管钙化指数多参数累积计算方法对下肢血管CT图像处理过程中，是先通过对下肢血管CT图像进行超像素分割，一方面可以将像素聚合在一起，构成具有形状规则和局部结构一致的多个个子区域块，实现了图像局部特征和结构信息整体表达，避免数据过多，提高处理速度，同时利用超像素降低了数据维度。另一方面，通过线性迭代聚类算法计算所以超像素区域内特征的平均值替换该区域中的像素值，可以最大限度地保留有效信息，减少噪声。
2、本发明通过Lab颜色空间提取钙化斑点所在超像素区域的亮度特征值，先是对原始图像L 0逐层使用低通滤波器以及子采样操作，可以得到原始图像L 0空间尺度变换的亮度强度图，可以加强图像的显著区域的边缘。再根据超像素区域内亮度最大像素点求得该超像素区域内所有像素灰度值的和，利用像素灰度值积分可以从Lab颜色空间中获得CT亮度值。
3、在提取CT图像中的钙化面积时，采用Canny算子对图像中钙化斑点进行边缘检测和轮廓提取，利用高斯滤波器对图像进行预处理了减少噪声影响；采用非极大值抑制让边缘 有一个准确的响应，在正确的位置检测边缘，提高了目标的准确性。用滞后阈值处理检测来连接边缘点，去除虚假边缘，使得边缘的定位精度提高。
4、本发明采用主动轮廓模型优化椭圆轮廓，即使用snake模型，在内力和外力的作用下使得轮廓变形，外部能量吸引活动轮廓不断逼近目标轮廓，最终演化结果被接收为目标轮廓，经过snake模型曲线演化并重新拟合后的轮廓更加逼近真实轮廓。
5、本发明基于聚类算法的下肢血管钙化指数多参数累积计算方法，另辟蹊径的通过对下肢血管的CT图像处理得到的CT亮度特征值、钙化面积，结合血管流体力学研究获得的相应的累积校正系数，给出了下肢血管钙化指数的量化数值，该下肢血管钙化指数的数值能够很大程度的表征下肢血管的钙化程度，为后续的糖尿病足面临截肢的诊断风险提供数据依据，有效降低经验判断带来的误差。此外，通过对CT图像进行处理获得钙化程度，避免人为判断因素的影响，提高了准确性。
附图说明
图1是本发明方法处理流程图；
图2是划分超像素区域流程图；
图3是糖尿病患者下肢CT图像；
图4是动脉血管中膜钙化和内膜钙化示意图；
图5是动脉粥样硬化的进程示意图。
具体实施方式
为了使本发明的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本发明进行进一步详细说明。应当理解，此处所描述的具体实施例仅用于解释本发明，并不用于限定本发明。
步骤1，使用CT等检测仪器对糖尿病患者下肢进行下肢血管CT图像的采集，得到需要进行分析的下肢血管CT图像；
步骤2，本申请利用线性迭代聚类算法(SLIC)将CT图像中钙化斑点均匀的分割在每一个超像素区域内；过程如下：
步骤2.1，设原始下肢血管CT图像具有X个像素，设置初始化超像素的数目为K，对原始CT图像进行均匀分割，此时分割出的每个超像素区域都有
Figure PCTCN2021094611-appb-000010
个像素；
步骤2.2，预设初始的聚类中心C和初始聚类中心C之间的间隔；
步骤2.3，基于欧几里距离搜寻聚类中心C领域内靠近C的像素点，并将这些像素点 划为一类，直到将每个聚类中心C所在领域内全部的像素点都划分结束；
步骤2.4，计算K个超像素区域内所有像元的平均特征向量值，根据平均特征向量值进行下一次聚类，迭代更新聚类中心，再次迭代直到迭代结束。
步骤2.5，对完成迭代后的超像素进行分割，获得超像素区域。
步骤3，完成超像素分割后，通过Lab颜色空间提取钙化斑点所在超像素区域的亮度特征值，具体过程为：
步骤3.1，提取出Lab中的亮度通道L，表征下肢血管CT图像亮度特征
Figure PCTCN2021094611-appb-000011
进而获得下肢血管CT图像的亮度图像L 0，Y为的中间变量，Y 0为是CIE标准定义白色的灰度值，f为矫正函数；
步骤3.2，基于L通道亮度图L 0提取钙化斑点所在超像素区域的亮度特征值；
步骤3.2.1，读取L通道亮度图像L 0，将L 0作为原始图像，通过高斯滤波1/2下采样得到图像L 1，L 1＝subsample(lpfilter(L 0))，其中，subsample()为下采样函数；lpfilter()为频域滤波函数，；对L 0使用频域滤波和下采样操作后能够获得L 0亮度强度后的图像L 1，经过该处理后图像的显著区域的边缘被有效增强；
步骤3.2.2，提取亮度增强后图像L 1内所有超像素区域内亮度最大像素点A(x,y)，并求得该超像素区域内所有像素灰度值的和；即：
Figure PCTCN2021094611-appb-000012
其中，(x,y)为像素坐标，y∈(y 1,y 2)，x∈(x 1,x 2)，y 1、y 2分别是该超像素区域内y轴方向的坐标值，x 1、x 2分别是该超像素区域内 x轴方向的坐标值，f(x,y)为(x,y)处的像素值，P y(x)为在横坐标在 x处的列向量像素灰度值累加和，再通过
Figure PCTCN2021094611-appb-000013
得到整个超像素区域内所有像素灰度值累加和；
步骤3.2.3，整个超像素区域内所有像素灰度值的累加和即为CT亮度值。
步骤4，提取CT图像中的钙化面积，由CT图像中可以看出，在患者下肢分布的钙化斑点呈现为外围不光滑的斑点状，因此，从图像中提取钙化面积需要进行以下处理：
步骤4.1，利用边缘检测和轮廓提取对钙化斑点处理的过程法为：
步骤4.1.1，首先，利用高斯滤波器对图像进行预处理，可以平滑图像，同时过滤噪声
步骤4.1.2，使用sobel算子计算滤波后图像中每个像素点的梯度大小和方向；
步骤4.1.3，基于像素的梯度强度对比选择边缘点，当某一像素的梯度强度大于沿正负梯度方向上的另外两个像素，则保留其为边缘点，反之则抑制该像素，即将该点抑制为0；通过上述选择过程能够使边缘点产生准确的响应，提高了边缘点提取的准确性。
步骤4.1.4，设定上限阈值，将上一步得到的边缘点与上限阈值相比较，进而对边缘点进行筛选；若上限阈值小于边缘点，则保留该点并设定改点为第一个边缘点；；接着查找该点的相邻点是否大于上限阈值，重复这一个过程，并将所有大于上限阈值的点进行连接，通过滞后阈值处理可以去除虚假边缘，使得边缘的定位精度提高。
步骤4.2，对分段椭圆进行拟合的方法为：
步骤4.2.1，将上步得到的轮廓随机分为n段，n∈[8，12]，n为该范围内的偶数；
步骤4.2.2，在每一段轮廓中任意选取12个不重复的点，利用最小二乘法拟合得到n个候选椭圆；
步骤4.2.3，设置判断阈值l 0，将点(x i,y i)与候选椭圆轮廓之间的距离l i与判断阈值l 0进行比较，若l i>l 0，则丢弃改点，不记票数，若l i≤l 0，则保留改点，记1票，并将该点的相关参数汇总得到数据集V i＝(x ic,y ic,a i,b i,θ i,n i,s i)，其中，候选椭圆的圆心为(x ic,y ic)，半长轴为a i，半短轴为b i，旋转角为θ i，s i代表每个段的序列号，n i为轮廓的段数；重复上述比较过程直至所有候选椭圆轮廓上的点都被比较过，将所有被保留点的M个的数据集汇总得到V＝{V i＝(x ic,y ic,a i,b i,θ i,n i,s i)|i＝1,2,......,M}，且记票数最多被认定为候选圆确定候选圆。本发明利用分段椭圆拟合能有效降低拟合错误椭圆的概率，提高准确性。
步骤4.3，采用主动轮廓模型对椭圆轮廓进行优化，最终获得钙化斑点的面积；
步骤4.3.1，利用snake模型在感兴趣区域的附近给出一条2D参数闭合曲线，通过最小化能量泛函，让这条闭合曲线在图像中发生变形并不断逼近目标轮廓，最终演化结果被接收为目标轮廓，轮廓曲线能量函数表示为：
Figure PCTCN2021094611-appb-000014
其中，E snake(v(s))是曲线能量，v(s)是snake轮廓的参数方程，E int是曲线的内部能量，它决定曲线的光滑度和连续性；E ext是外界给曲线的能量，它使得曲线朝着目标的特征方 向移动，s是描述边界的自变量。
步骤4.3.2，使用最小二乘圆拟合方法重新拟合圆，通过所有圆上的边缘点坐标的加权函数得到圆的中心，即得到钙化斑点的中心(X,Y)，其中，
Figure PCTCN2021094611-appb-000015
进而计算出钙化斑点的直径
Figure PCTCN2021094611-appb-000016
其中，x i和y i分别表示钙化斑点轮廓上某一点的坐标，N为钙化斑点轮廓上点的个数；最终的得出得到钙化斑点的面积。
步骤5，将上述步骤3和步骤4分别得到的CT亮度值ρ和钙化斑点的面积S代入Cal＝k·ρ·S获得CT图像中钙化程度的判断值，其中，k为累积校正系数，钙化指数除了与CT图像的钙化斑点的面积S、钙化斑点的亮度，以及狭窄度有关联，还和壁面剪应力w、震荡剪切指数o、壁面压力p关系密切，截取钙化斑点所在区域的4cm长的血管；本发明利用累积校正系数k对钙化程度进行校正，因为在下肢动脉中的钙化类型包括内膜钙化和中膜钙化，而根据CT图像中的钙化斑点，是无法直接区分出是内膜钙化还是中膜钙化如图4和5，但是内膜钙化容易引起动脉管腔的狭窄，当动脉管腔出现血管狭窄之后，血流的动力学在狭窄段明显变化；因此在本发明中通过增加累积校正系数k对CT亮度值ρ和钙化斑点的面积S进行综合计算，进而获得更为准确的钙化程度的判断值，根据Cal＝k·ρ·S获得CT图像中钙化指数的判断值。
其中，下肢动脉钙化斑块的面积S累积计算可以通过利用分段椭圆对处理后的图像中钙化斑点进行拟合并进行优化处理得到钙化斑点的半径，进而计算出钙化斑点的面积直接获得；或者通过评分标准直接给出：由下肢动脉前壁和后壁钙化斑块面积的大小，分为四个等级：Ⅰ、无钙化，评0分；Ⅱ、钙化范围小于1/3的动脉壁长度，评1分；Ⅲ、钙化范围累及1/3～2/3的动脉壁长度，评2分；Ⅳ、钙化范围大于2/3的动脉壁长度，评3分；故进行血管分段处理，将下肢血管以小腿正中为分界线，全部分为上下两段，包括：胫后动脉近端、胫后动脉远端、胫前动脉近端、胫前动脉远端、腓动脉近端、腓动脉远端。
上述钙化斑点所在超像素区域的亮度值ρ可以直接通过聚类算法中的Lab颜色空间提取钙化斑点所在超像素区域的CT亮度特征值获得。
上述累积校正系数可以按照如下计算公式获取：k＝k h*k w*k o*k p，k h为狭窄度累积评分；k w为下肢动脉血管壁面剪应力w累积评分；k o为下肢动脉血管震荡剪切指数o累积评分；k p为下肢动脉血管壁面压力p累积评分；下面是具体的累积校正系数k的计算标准：
下肢动脉血管狭窄度的累积计算标准为：根据下肢动脉血管的管腔面积，分为四个等 级：Ⅰ、管腔直径缩小1％-25％，狭窄度累积评分k h评1分，Ⅱ、管腔直径缩小25％-50％，狭窄度累积评分k h评2分，Ⅲ、管腔直径缩小51％-75％，狭窄度累积评分k h评3分，Ⅳ、管腔直径缩小76％-100％，狭窄度累积评分k h评4分；狭窄度数据的采集方法为：采集下肢动脉在有钙化斑块段的血管狭窄度的最大值h max和狭窄度的平均值
Figure PCTCN2021094611-appb-000017
根据
Figure PCTCN2021094611-appb-000018
确定该段的血管狭窄度
Figure PCTCN2021094611-appb-000019
其中a、b为常系数。
下肢动脉血管壁面剪应力w累积计算标准为：动脉系统中的壁面剪应力一般为(10～70)dynes/cm 2，壁面剪应力w为(0～4)dynes/cm 2时累积评分k w为3分，壁面剪应力w为(5～10)dynes/cm 2时累积评分k w为2分，壁面剪应力w为(11～70)dynes/cm 2时累积评分k w为1分。
下肢动脉血管震荡剪切指数o累积计算标准为：指标正常范围：0～0.5，震荡剪切指数低于0.2时累积评分k o为3分，震荡剪切指数介于0.2～0.3时累积评分k o为2分，震荡剪切指数介于0.3～0.5时累积评分k o为1分。
下肢动脉血管壁面压力p累积计算标准为：指标正常范围：收缩压：90～140mmHg，舒张压：60～90mmHg；收缩压低于90mmHg、舒张压低于60mmHg累积评分k p为3分，收缩压介于90～140mmHg、舒张压介于60～90mmHg累积评分k p为2分，收缩压高于140mmHg、舒张压高于90mmHg累积评分k p为1分。
本发明的基于聚类算法的下肢血管钙化指数多参数累积计算方法，直接目的并不能获得糖尿病足疾病的诊断结果或健康状况，而只是从活的人体获取作为中间结果的信息的方法，为糖尿病足的疾病诊断提供中间数据支撑，不属于疾病诊断的范畴。
以上实施例仅用于说明本发明的设计思想和特点，其目的在于使本领域内的技术人员能够了解本发明的内容并据以实施，本发明的保护范围不限于上述实施例。所以，凡依据本发明所揭示的原理、设计思路所作的等同变化或修饰，均在本发明的保护范围之内。

Claims (8)

  1. 基于聚类算法的下肢血管钙化指数多参数累积计算方法，其特征在于，包括如下步骤：
    步骤1，采集待分析的下肢血管CT图像；
    步骤2，使用线性迭代聚类算法将CT图像中钙化斑点均匀的分割在每一个超像素区域内；
    步骤3，完成超像素分割后，通过Lab颜色空间提取钙化斑点所在超像素区域的CT亮度特征值；
    步骤4，对CT图像中的钙化斑点进行边缘检测和轮廓提取，利用分段椭圆对处理后的图像中钙化斑点进行拟合并进行优化处理得到钙化斑点的半径，进而计算出钙化斑点的面积；
    步骤5，根据Cal＝k·ρ·S获得CT图像中钙化程度的判断值，其中，ρ为CT亮度特征值，S为钙化斑点的面积，k为累积校正系数。
  2. 根据权利要求1所述的基于聚类算法的下肢血管钙化指数多参数累积计算方法，其特征在于，所述步骤2将CT图像进行超像素分割的方法为：
    步骤2.1，对所采集的原始下肢血管CT图像进行均匀的超像素分割，设原始下肢血管CT图像中的像素共有X个，原始下肢血管CT图像被分割成K个，则每个超像素都有
    Figure PCTCN2021094611-appb-100001
    个像素；
    步骤2.2，预设初始的聚类中心C和初始聚类中心C之间的间隔；
    步骤2.3，基于欧几里距离搜寻聚类中心C领域内靠近C的像素点，并将这些像素点划为一类；
    步骤2.4，计算K个超像素区域内所有像元的平均特征向量值，根据平均特征向量值进行下一次聚类，迭代更新聚类中心，再次迭代直到迭代结束；
    步骤2.5，对完成迭代后的超像素进行分割，获得超像素区域。
  3. 根据权利要求1所述的基于聚类算法的下肢血管钙化指数多参数累积计算方法，其特征在于，所述步骤3中提取钙化斑点所在超像素区域的亮度特征值的方法为：
    步骤3.1，提取出Lab中的亮度通道L，表征下肢血管CT图像亮度特征
    Figure PCTCN2021094611-appb-100002
    进而获得下肢血管CT图像的亮度图像L 0，Y为的中间变量，Y 0为是CIE标准定义白色的灰度值，f为矫正函数；
    步骤3.2，基于亮度通道L的亮度图L 0提取钙化斑点所在超像素区域的亮度特征值。
  4. 根据权利要求3所述的基于聚类算法的下肢血管钙化指数多参数累积计算方法，其特征在于，步骤3.2提取亮度特征值的过程为：
    步骤3.2.1，使用高斯滤波1/2下采样对L 0进行处理后得到图像L 1，L 1＝subsample(lpfilter(L 0))，其中，subsample()为下采样函数，lpfilter()为频域滤波函数；
    步骤3.2.2，提取经高斯滤波后的超像素区域内亮度最大像素点A(x,y)，并求得该超像素区域内所有像素灰度值的和
    Figure PCTCN2021094611-appb-100003
    其中，(x,y)为像素坐标，y∈(y 1,y 2)，x∈(x 1,x 2)，y 1、y 2分别是该超像素区域内y轴方向的坐标值，x 1、x 2分别是该超像素区域内x轴方向的坐标值，f(x,y)为(x,y)处的像素值，P y(x)为在横坐标在x处的列向量像素灰度值累加和，再通过
    Figure PCTCN2021094611-appb-100004
    得到整个超像素区域内所有像素灰度值累加和；
    步骤3.2.3，整个超像素区域内所有像素灰度值的累加和即为CT亮度值。
  5. 根据权利要求1所述的基于聚类算法的下肢血管钙化指数多参数累积计算方法，其特征在于，利用边缘检测和轮廓提取对钙化斑点处理的过程法为：
    步骤4.1.1，利用高斯滤波器对图像进行预处理；
    步骤4.1.2，使用sobel算子计算滤波后图像中每个像素点的梯度大小和方向；
    步骤4.1.3，基于像素的梯度强度对比选择边缘点，当某一像素的梯度强度大于沿正负梯度方向上的另外两个像素，则保留其为边缘点，反之则抑制该像素；
    步骤4.1.4，将上一步得到的边缘点与上限阈值相比较，进而对边缘点进行筛选；若上限阈值小于边缘点，则保留该点并设定改点为第一个边缘点；接着查找该点的相邻点是否小于上限阈值，重复这一个过程，并将所有大于上限阈值的点进行连接。
  6. 根据权利要求5所述的基于聚类算法的下肢血管钙化指数多参数累积计算方法，其特征在于，所述分段椭圆进行拟合的方法为：
    步骤4.2.1，将得到的轮廓随机分为n段；
    步骤4.2.2，在每一段轮廓中任意选取12个不重复的点，利用最小二乘法拟合得到n个候选椭圆；
    步骤4.2.3，设置判断阈值l 0，将点(x i,y i)与候选椭圆轮廓之间的距离l i与判断阈值l 0进行比较，若l i>l 0，则丢弃改点，不记票数，若l i≤l 0，则保留改点，记1票，并将该点的相关参数汇总得到数据集V i＝(x ic,y ic,a i,b i,θ i,n i,s i)，其中，候选椭圆的圆心为(x ic,y ic)，半长轴为a i，半短轴为b i，旋转角为θ i，s i代表每个段的序列号，n i为轮廓的段数；重复上述比较过程直至所有候选椭圆轮廓上的点都被比较过，将所有被保留的M个点的数据集汇总得到V＝{V i＝(x ic,y ic,a i,b i,θ i,n i,s i)|i＝1,2,......,M}，且记票数最多被认定为候选圆。
  7. 根据权利要求6所述的基于聚类算法的下肢血管钙化指数多参数累积计算方法，其特征在于，步骤4.3.1，利用snake模型在感兴趣区域附近给出一条2D参数闭合曲线，通过最小化能量泛函，让这条闭合曲线在图像中发生变形并不断逼近目标轮廓，最终演化结果被接收为目标轮廓，轮廓曲线能量函数表示为：
    Figure PCTCN2021094611-appb-100005
    其中，E snake(v(s))是曲线能量，v(s)是snake轮廓的参数方程，E int是曲线的内部能量，它决定曲线的光滑度和连续性；E ext是外界给曲线的能量，它使得曲线朝着目标的特征方向移动，s是描述边界的自变量；
    步骤4.3.2，使用最小二乘圆拟合方法重新拟合圆，通过所有圆上的边缘点坐标的加权函数得到圆的中心，即得到钙化斑点的中心(X,Y)，其中，
    Figure PCTCN2021094611-appb-100006
    进而计算出钙化斑点的直径
    Figure PCTCN2021094611-appb-100007
    其中，x i和y i分别表示钙化斑点轮廓上某一点的坐标，N为钙化斑点轮廓上点的个数；最终的得出得到钙化斑点的面积。
  8. 根据权利要求1所述的基于聚类算法的下肢血管钙化指数多参数累积计算方法，其特征在于，所述累积校正系数k＝k h*k w*k o*k p，其中，k h为狭窄度累积评分；k w为下肢动脉血管壁面剪应力w累积评分；k o为下肢动脉血管震荡剪切指数o累积评分；k p为下肢动脉血管壁面压力p累积评分；
    下肢动脉血管狭窄度的累积计算标准为：根据下肢动脉血管的管腔面积，分为四个等 级：Ⅰ、管腔直径缩小1％-25％，狭窄度累积评分k h评1分，Ⅱ、管腔直径缩小25％-50％，狭窄度累积评分k h评2分，Ⅲ、管腔直径缩小51％-75％，狭窄度累积评分k h评3分，Ⅳ、管腔直径缩小76％-100％，狭窄度累积评分k h评4分；狭窄度数据的采集方法为：采集下肢动脉在有钙化斑块段的血管狭窄度的最大值h max和狭窄度的平均值
    Figure PCTCN2021094611-appb-100008
    根据
    Figure PCTCN2021094611-appb-100009
    确定该段的血管狭窄度
    Figure PCTCN2021094611-appb-100010
    其中a、b为常系数；
    下肢动脉血管壁面剪应力w累积计算标准为：动脉系统中的壁面剪应力一般为10～70dynes/cm 2，壁面剪应力w为0～4dynes/cm 2时累积评分k w为3分，壁面剪应力w为5～10dynes/cm 2时累积评分k w为2分，壁面剪应力w为11～70dynes/cm 2时累积评分k w为1分；
    下肢动脉血管震荡剪切指数o累积计算标准为：指标正常范围：0～0.5，震荡剪切指数低于0.2时累积评分k o为3分，震荡剪切指数介于0.2～0.3时累积评分k o为2分，震荡剪切指数介于0.3～0.5时累积评分k o为1分；
    下肢动脉血管壁面压力p累积计算标准为：指标正常范围：收缩压：90～140mmHg，舒张压：60～90mmHg；收缩压低于90mmHg、舒张压低于60mmHg累积评分k p为3分，收缩压介于90～140mmHg、舒张压介于60～90mmHg累积评分k p为2分，收缩压高于140mmHg、舒张压高于90mmHg累积评分k p为1分。
PCT/CN2021/094611 2020-05-28 2021-05-19 基于聚类算法的下肢血管钙化指数多参数累积计算方法 WO2021238739A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US17/432,094 US11557072B2 (en) 2020-05-28 2021-05-19 Clustering algorithm-based multi-parameter cumulative calculation method for lower limb vascular calcification indexes

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010466288.7A CN111667467B (zh) 2020-05-28 2020-05-28 基于聚类算法的下肢血管钙化指数多参数累积计算方法
CN202010466288.7 2020-05-28

Publications (1)

Publication Number Publication Date
WO2021238739A1 true WO2021238739A1 (zh) 2021-12-02

Family

ID=72384885

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/094611 WO2021238739A1 (zh) 2020-05-28 2021-05-19 基于聚类算法的下肢血管钙化指数多参数累积计算方法

Country Status (3)

Country Link
US (1) US11557072B2 (zh)
CN (1) CN111667467B (zh)
WO (1) WO2021238739A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114366164A (zh) * 2022-01-18 2022-04-19 河南科技大学 一种动脉粥样硬化斑块超声定量评价与风险评估方法
CN115029704A (zh) * 2022-08-09 2022-09-09 江苏贯森新材料科技有限公司 一种用于不锈钢酸洗工艺的智能控制方法及装置

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111667467B (zh) 2020-05-28 2021-01-26 江苏大学附属医院 基于聚类算法的下肢血管钙化指数多参数累积计算方法
CN112489806B (zh) * 2020-11-27 2022-02-15 南通市第一人民医院 一种糖尿病足患者病状信息智能化管理方法及系统
CN115393318A (zh) * 2022-08-25 2022-11-25 朗天药业(湖北)有限公司 一种血塞通滴丸外观质量检测方法及系统
CN116109662B (zh) * 2023-04-13 2023-06-23 中国科学院国家空间科学中心 一种红外图像的超像素分割方法
CN117237389B (zh) * 2023-11-14 2024-01-19 深圳市亿康医疗技术有限公司 一种中耳胆脂瘤ct影像分割方法
CN118154615B (zh) * 2024-05-13 2024-07-19 山东聚宁机械有限公司 一种挖掘机履带板的板体质量智能检测方法
CN118334656B (zh) * 2024-06-12 2024-08-20 济南市计量检定测试院 基于机器视觉的计量检测数据识别方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080044070A1 (en) * 2006-08-16 2008-02-21 Fuji Photo Film Co., Ltd. Abnormal mass candidate detecting apparatus, method and computer-readable medium
CN103745227A (zh) * 2013-12-31 2014-04-23 沈阳航空航天大学 一种基于多维信息的肺结节良恶性鉴别方法
CN106447645A (zh) * 2016-04-05 2017-02-22 天津大学 增强ct图像中冠脉钙化检测及量化装置和方法
CN111667467A (zh) * 2020-05-28 2020-09-15 江苏大学附属医院 基于聚类算法的下肢血管钙化指数多参数累积计算方法

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7865000B2 (en) * 2004-06-10 2011-01-04 Kurume University Medical image processing apparatus for analyzing properties of living body based on image data thereof
CN103337096B (zh) * 2013-07-19 2015-09-09 东南大学 一种冠状动脉ct造影图像钙化点检测方法
EP3985557B1 (en) * 2016-09-23 2023-11-08 Curemetrix, Inc. Mapping of breast arterial calcifications
CN110021016B (zh) * 2019-04-01 2020-12-18 数坤（北京）网络科技有限公司 一种钙化检测方法
CN110600126A (zh) * 2019-09-19 2019-12-20 江苏大学附属医院 一种糖尿病足下肢动脉钙化斑块图像的辅助评价方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080044070A1 (en) * 2006-08-16 2008-02-21 Fuji Photo Film Co., Ltd. Abnormal mass candidate detecting apparatus, method and computer-readable medium
CN103745227A (zh) * 2013-12-31 2014-04-23 沈阳航空航天大学 一种基于多维信息的肺结节良恶性鉴别方法
CN106447645A (zh) * 2016-04-05 2017-02-22 天津大学 增强ct图像中冠脉钙化检测及量化装置和方法
CN111667467A (zh) * 2020-05-28 2020-09-15 江苏大学附属医院 基于聚类算法的下肢血管钙化指数多参数累积计算方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
THE SNAIL THAT COLLECTS MEDICINAL HERBS: "Calcium Score Calculation in Medical Image Processing", CNBLOG.COM, 12 April 2016 (2016-04-12), XP055872252, Retrieved from the Internet <URL:https://www.cnblogs.com/mothe123/p/4481626.html> *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114366164A (zh) * 2022-01-18 2022-04-19 河南科技大学 一种动脉粥样硬化斑块超声定量评价与风险评估方法
CN114366164B (zh) * 2022-01-18 2024-05-03 河南科技大学 一种动脉粥样硬化斑块超声定量评价与风险评估方法
CN115029704A (zh) * 2022-08-09 2022-09-09 江苏贯森新材料科技有限公司 一种用于不锈钢酸洗工艺的智能控制方法及装置
CN115029704B (zh) * 2022-08-09 2022-12-09 江苏贯森新材料科技有限公司 一种用于不锈钢酸洗工艺的智能控制方法及装置

Also Published As

Publication number Publication date
US20220148237A1 (en) 2022-05-12
US11557072B2 (en) 2023-01-17
CN111667467A (zh) 2020-09-15
CN111667467B (zh) 2021-01-26

Similar Documents

Publication Publication Date Title
WO2021238739A1 (zh) 基于聚类算法的下肢血管钙化指数多参数累积计算方法
CN110969622B (zh) 用于辅助肺炎诊断的图像处理方法和系统
Liu et al. Automatic whole heart segmentation using a two-stage u-net framework and an adaptive threshold window
CN108618749B (zh) 基于便携式数字化眼底照相机的视网膜血管三维重建方法
CN110448319A (zh) 基于造影影像及冠状动脉的血流速度计算方法
CN113420826B (zh) 一种肝脏病灶图像处理系统及图像处理方法
WO2022105623A1 (zh) 一种基于迁移学习的颅内血管病灶识别方法
CN111080700A (zh) 医疗器械图像检测方法及装置
CN116993628B (zh) 一种用于肿瘤射频消融引导的ct图像增强系统
WO2023151280A1 (zh) 双模冠脉血管图像三维融合的方法和融合系统
CN112562058B (zh) 一种基于迁移学习的颅内血管模拟三维模型快速建立方法
CN113470137A (zh) 基于灰度加权的ivoct图像导丝伪影去除方法
CN110353639B (zh) 一种基于血管增强造影的供血区定量方法及系统
CN114332910A (zh) 一种面向远红外图像的相似特征计算的人体部位分割方法
CN111292285B (zh) 一种基于朴素贝叶斯与支持向量机的糖网病自动筛查方法
LU102573B1 (en) Clustering Algorithm-based Multi-parameter Cumulative Calculation Method for Lower Limb Vascular Calcification Indexes
CN114266742A (zh) 脑血管cta图像的狭窄区域检测方法
Hamou et al. Carotid ultrasound segmentation using DP active contours
JP2022052210A (ja) 情報処理装置、情報処理方法及びプログラム
Mi et al. Detecting carotid intima-media from small-sample ultrasound images
Rajeoni et al. Vascular system segmentation using deep learning
Zhang et al. Diagnosis of Coronary Heart Disease by Optical Coherence Tomography Using Random Walk Algorithm
CN112037242B (zh) 皮肤光学相干层析成像图像中表皮层的自动识别方法
Wong-od et al. Intravascular ultrasound image recovery and segmentation based on circular analysis
CN112669439B (zh) 基于迁移学习的颅内血管造影增强三维模型的建立方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21812113

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21812113

Country of ref document: EP

Kind code of ref document: A1