技术领域technical field
本发明涉及医疗器械、图像处理技术，具体讲,涉及基于水平集和动态规划动脉内中膜厚度测量装置和算法。The invention relates to medical equipment and image processing technology, in particular to a device and algorithm for measuring arterial intima-media thickness based on level set and dynamic programming.
背景技术Background technique
心脑血管疾病(Cardiovascular diseases,CVDs)是人类身体健康的首要疾病，已经成为人类生命第一杀手。而动脉粥状硬化最常见的病变部位就是颈总动脉(CommonCarotid Artery，CCA)，血液中的脂质随着心脑血管疾病的发展在血管内壁沉积，从而使内中膜结构(Intima-Media Complex，IMC)变厚增生，动脉变窄，内中膜厚度(Intima-MediaThickness，IMT)如果增大到一定程度，血流速度将会减慢，使大脑供血不畅，大脑功能也受到影响。因此，心脑血管疾病的早期病变表现在内中膜增厚上，内中膜厚度是预测心脑血管疾病严重程度的重要指标。临床上，内中膜厚度通常是医生本人对颈总动脉超声图像进行手工定点或边界描绘得出，主观性强，更加依赖个人经验。因此，本发明提出了一种结合水平集分割和动态规划的颈动脉超声图像内中膜厚度测量算法，利用计算机代替人工进行边界膜自动分割并计算IMT相关参数，以减少人工耗时和工作量，并降低测量差异，为心脑血管疾病早期诊断提供准确的依据。Cardiovascular diseases (CVDs) are the primary diseases of human health and have become the number one killer of human life. The most common lesion site of atherosclerosis is the Common Carotid Artery (CCA), and the lipids in the blood deposit on the inner wall of the blood vessel with the development of cardiovascular and cerebrovascular diseases, so that the Intima-Media Complex (Intima-Media Complex) , IMC) thickening and hyperplasia, arterial narrowing, if the Intima-Media Thickness (IMT) increases to a certain extent, the blood flow velocity will slow down, making the blood supply to the brain poor, and brain function will also be affected. Therefore, early lesions of cardiovascular and cerebrovascular diseases are manifested in intima-media thickening, and intima-media thickness is an important indicator for predicting the severity of cardiovascular and cerebrovascular diseases. Clinically, the intima-media thickness is usually determined by the doctor himself on the ultrasound images of the common carotid artery by manually fixing points or delineating the boundaries, which is highly subjective and more dependent on personal experience. Therefore, the present invention proposes a carotid artery ultrasound image intima-media thickness measurement algorithm that combines level set segmentation and dynamic programming, and uses a computer to replace manual boundary membrane automatic segmentation and calculate IMT-related parameters to reduce manual time-consuming and workload , and reduce measurement differences, providing an accurate basis for early diagnosis of cardiovascular and cerebrovascular diseases.
基于几何变形模型的水平集方法由Osher和Sethian于1982年提出。采用水平集分割方法所表示的几何曲线演化模型理论应用于图像分割领域的基本原理是:利用速度函数F(包含曲线变形力，比如曲率驱动力或者法向常量力或),在图像的空间范围对水平集函数演化,致使水平集函数的第零水平集所描述的闭合曲线能够在目标轮廓线所在的位置停止。水平集方法的思想中最重要的就是Hamilton-Jacobi方程，它为运动的隐式曲面进行基于时间的方程的数值求解该方法的基本原理是，将曲线或曲面作为零水平集嵌入到高一维的水平集函数中，通过一个更高维的函数来表达低维的曲线或曲面的演化过程。用水平集函数的零水平集函数表示闭合曲线的描述方法的核心思想是:把演化的闭合曲线C＝{(x(p),y(p)),p∈[0,1]}嵌入到2D水平集函数φ(x,y)中,用水平集函数的第零水平集隐式地表示闭合曲线C:C＝{(x,y),φ(x,y)＝0}。此时曲线的法向量切向量并且在第t时刻有The level set method based on geometric deformation model was proposed by Osher and Sethian in 1982. The basic principle of applying the geometric curve evolution model theory represented by the level set segmentation method to the field of image segmentation is: using the velocity function F (including curve deformation force, such as curvature driving force or normal constant force or), in the spatial range of the image For the evolution of the level set function, the closed curve described by the zeroth level set of the level set function can stop at the position of the target contour. The most important idea of the level set method is the Hamilton-Jacobi equation, which performs the numerical solution of the time-based equation for the implicit surface of motion. The basic principle of the method is to embed the curve or surface as a zero level set into a higher dimension In the level set function of , a higher-dimensional function is used to express the evolution process of a low-dimensional curve or surface. The core idea of the closed curve description method using the zero level set function of the level set function is to embed the evolved closed curve C={(x(p),y(p)),p∈[0,1]} into In the 2D level set function φ(x, y), the closed curve C is represented implicitly by the zeroth level set of the level set function: C={(x, y), φ(x, y)=0}. The normal vector of the curve at this time Tangent vector And at time t there is
φ(C(p,t),t)＝0 (1)φ(C(p,t),t)=0 (1)
两边求导可得到Derivation on both sides gives
代入式(1)的曲线演化方程以及法向量可得到水平集演化方程Substituting the curve evolution equation and normal vector into formula (1) The level set evolution equation can be obtained
其中F为水平集函数演化的速度函数,同样地根据其取值的不同也存在不同的演化方式。通过速度函数F的作用不断演化水平集函数,最后第零水平集函数所形成的轮廓就是曲线演化的最终形式。Among them, F is the speed function of level set function evolution, and there are different evolution modes according to its different values. The level set function is continuously evolved through the action of the velocity function F, and finally the contour formed by the zeroth level set function is the final form of the curve evolution.
动态规划(Dynamic Programming)是一种高效的优化方法，可以解决函数自变量非全相关情况下的优化问题。动态规划技术的关键在于定义一个代价函数,该代价函数包括边缘强度、边界光滑等约束,使代价函数达到极小能提取从左到右的边界它将复杂问题分解为一系列简单的子问题，并通过迭代的方法得到问题的解。对于下面这个问题：Dynamic programming (Dynamic Programming) is an efficient optimization method that can solve the optimization problem in the case of non-fully correlated function independent variables. The key to dynamic programming technology is to define a cost function, which includes constraints such as edge strength and boundary smoothness, so that the cost function can be minimized and the boundary from left to right can be extracted. It decomposes complex problems into a series of simple sub-problems, And get the solution of the problem through iterative method. For this question:
h(x1,x2,x3,x4)＝h1(x1,x2)+h2(x2,x3)+h3(x3,x4) (4)h(x1 ,x2 ,x3 ,x4 )=h1 (x1 ,x2 )+h2 (x2 ,x3 )+h3 (x3 ,x4 ) (4)
求解(x1,x2,x3,x4)，使得h(x1,x2,x3,x4)取得最大值。Solve (x1 ,x2 ,x3 ,x4 ) so that h(x1 ,x2 ,x3 ,x4 ) obtains the maximum value.
利用动态规划算法解决这个问题时可以分为以下四步：Using dynamic programming algorithm to solve this problem can be divided into the following four steps:
第一步，对于每一个x2计算f1(x2)的最大值并保存对应的x1与f1(x2)值，即The first step is to calculate the maximum value of f1 (x2 ) for each x2 and save the corresponding values of x1 and f1 (x2 ), namely
第二步，对于每一个x3计算f2(x3)的最大值并保存对应的x2与f2(x3)值，即The second step is to calculate the maximum value of f2 (x3 ) for each x3 and save the corresponding values of x2 and f2 (x3 ), namely
第三步，对于每一个x4计算f3(x4)的最大值并保存对应的x3与f3(x4)值，即The third step is to calculate the maximum value of f3 (x4 ) for each x4 and save the corresponding values of x3 and f3 (x4 ), namely
第四步，选出x4使得f3(x4)值达到最大，并由“逆推法”选出与该f3(x4)对应的x3、与f2(x3)对应的x2和与f1(x2)对应的x1。The fourth step is to select x4 so that the value of f3 (x4 ) reaches the maximum value, and select x3 corresponding to f3 (x4 ) and f2 (x3 ) corresponding to x2 and x1 corresponding to f1 (x2 ).
最终求解出满足条件的(x1,x2,x3,x4)。Finally, (x1 , x2 , x3 , x4 ) that satisfy the conditions are solved.
动态规划求解的优点是计算速度快。对于含有N个自变量(x1,x2,...,xN)、每个自变量有K种取值的h(x1,x2,...,xN)优化问题，动态规划算法的计算量为(N-1)K2+N。对于同样规模的问题，若用贪心算法求解的计算量为KN。The advantage of dynamic programming solution is fast calculation speed. For h(x1 ,x2 ,...,xN ) optimization problem with N independent variables (x1 ,x2 ,...,xN ) and each independent variable has K values, the dynamic The calculation amount of the planning algorithm is (N-1)K2 +N. For problems of the same scale, if the greedy algorithm is used, the calculation amount is KN .
发明内容Contents of the invention
为克服现有技术的不足，本发明旨在将计算机图像处理技术应用在颈动脉超声波图像内中膜厚度测量中，避免手动测量方式存在的低效率及不稳定的缺陷，利用计算机算法提供快速、可靠的IMT参数，为心脑血管疾病早期诊断提供准确的依据。本发明采用的技术方案是，水平集动态规划颈总动脉内中膜厚度测量方法，包括如下步骤：In order to overcome the deficiencies of the prior art, the present invention aims to apply computer image processing technology to the measurement of intima-media thickness in carotid ultrasound images, avoid the defects of low efficiency and instability in manual measurement, and use computer algorithms to provide fast, Reliable IMT parameters provide accurate basis for early diagnosis of cardiovascular and cerebrovascular diseases. The technical solution adopted in the present invention is that the level set dynamic programming common carotid artery intima-media thickness measurement method comprises the following steps:
1)图像裁剪：裁剪移除超声图像中非测试信息部分，只保留血管部分；1) Image cropping: cropping removes the non-test information part of the ultrasound image, and only keeps the blood vessel part;
2)提取感兴趣区域ROI(Region of Interest)：自动提取ROI；2) Extract ROI (Region of Interest): automatically extract ROI;
3)图像动态拉伸：将图像灰度范围统一拉伸至0-255；3) Image dynamic stretching: Stretch the grayscale range of the image uniformly to 0-255;
4)图像滤波：采用双边滤波算法去除噪声，同时保留图像边缘信息；4) Image filtering: use bilateral filtering algorithm to remove noise while retaining image edge information;
5)估计管腔-内膜边界LII(Lumen-Intima Interface)：采用形态学操作避免内膜缺陷的干扰，使得估计的LII更加靠近真是边界；5) Estimation of lumen-intima boundary LII (Lumen-Intima Interface): Morphological operations are used to avoid the interference of intima defects, so that the estimated LII is closer to the real boundary;
6)估计中膜-外膜边界MAI(Media-Advantia Interface)：根据梯度图信息估计MAI；6) Estimating the Media-Advantia Interface MAI (Media-Advantia Interface): Estimate the MAI according to the gradient map information;
7)动态规划调整估计边界：以梯度值与曲率值的加权和作为动态规划的能量函数，进行迭代计算获得膜边界特性；7) Dynamic programming to adjust the estimated boundary: the weighted sum of the gradient value and the curvature value is used as the energy function of the dynamic programming, and iterative calculation is performed to obtain the membrane boundary characteristics;
8)后处理：计算并评估IMT，对自动测量的IMT值进行判别，避免测量结果没有实际意义。8) Post-processing: calculate and evaluate the IMT, and discriminate the automatically measured IMT value to prevent the measurement results from being meaningless.
提取ROI具体步骤是，首先直接对图像边缘使用连续曲线描述,与此同时采用图像信息对能量泛函进行定义，能量泛函的自变量含有边界轮廓曲线；接着采用动态Euler-Lagrange方程的格式得到这个能量泛函相对应的曲线演化的方程；最终采用水平集方法对沿着能量模拟初始曲线下降最快的方向的演化过程进行模拟，以求出最佳的边界轮廓曲线，提取出ROI部分。The specific steps of extracting ROI are as follows: firstly, describe the edge of the image directly using a continuous curve, and at the same time use image information to define the energy functional function, the independent variable of the energy functional function contains the boundary contour curve; then use the format of the dynamic Euler-Lagrange equation to obtain The equation of the curve evolution corresponding to this energy functional; finally, the level set method is used to simulate the evolution process along the direction of the fastest decline of the initial energy simulation curve, so as to find the best boundary contour curve and extract the ROI part.
图像动态拉伸具体步骤是，将高于m的灰度级进行压缩，直到它压缩到高灰度级的较窄范围内，而将低于m的灰度级进行压缩，使其存在于低灰度级的较窄范围内，拉伸中间的灰度级，增强各部分图像的反差，以使图像拥有比较高的对比度；拉伸公式如下：The specific steps of image dynamic stretching are to compress the gray level higher than m until it is compressed into a narrow range of high gray level, and compress the gray level lower than m to make it exist in the low range. Within a narrow range of gray levels, the middle gray level is stretched to enhance the contrast of each part of the image so that the image has a relatively high contrast; the stretching formula is as follows:
其中，s表示输出图像的灰度值，r表示输入图像灰度值，E是一个可以控制函数斜率的变量，m作为一个阈值可以被用户根据图像处理方面的实际需要进行参数设置。Among them, s represents the gray value of the output image, r represents the gray value of the input image, E is a variable that can control the slope of the function, and m is a threshold that can be set by the user according to the actual needs of image processing.
双边滤波公式如下：The bilateral filtering formula is as follows:
其中，为双边滤波的归一化常数，x,ξ表示像素点，f(x)表示x像素点的灰度值，滤波器为尺寸10×10的矩阵，表示像素ξ与x的灰度相似度，表示像素ξ与x的距离相似度，σ1为基于灰度相似性计算出的高斯标准差，基于距离相似性就算出的高斯函数标准差为σ2。in, is the normalization constant of bilateral filtering, x and ξ represent pixels, f(x) represents the gray value of x pixels, and the filter is a matrix of size 10×10, Indicates the gray similarity between pixel ξ and x, Indicates the distance similarity between pixel ξ and x, σ1 is the Gaussian standard deviation calculated based on the gray similarity, and the standard deviation of the Gaussian function calculated based on the distance similarity is σ2 .
估计LII步骤具体是，将ROI的灰度值归一化到区间[0,1]范围内，归一化函数的数学表达式如下：Specifically, the step of estimating LII is to normalize the gray value of the ROI to the interval [0,1]. The mathematical expression of the normalization function is as follows:
其中，g(x)为灰度实际值，fmin和fmax分别是ROI内灰度值的最小值和最大值；Among them, g(x) is the actual value of the gray scale, and fmin and fmax are the minimum and maximum values of the gray value in the ROI, respectively;
归一化后得到的直方图由3个峰和2个谷组成。假设2个谷最低点的灰度值为T1和T2，在ROI中沿竖直方向从上往下搜索，具体搜索步骤如下：The resulting histogram after normalization consists of 3 peaks and 2 valleys. Assuming that the gray values of the lowest points of the two valleys are T1 and T2 , search from top to bottom in the vertical direction in the ROI. The specific search steps are as follows:
(1)令j＝1；将i1从1增加到M进行搜索直到f(i1,j)≥T1；(i1,j)即所估计的LII起点；(1) Set j=1; increase i1 from 1 to M to search until f(i1 ,j)≥T1 ; (i1 ,j) is the estimated starting point of LII;
(2)将j增加为j+1，即在ROI的下一列搜索直到f(i2,j)≥T1；(2) Increase j to j+1, that is, search in the next column of ROI until f(i2 ,j)≥T1 ;
(3)若i2>i1，将(i1+1,j)作为估计LII上的一个新的点，若i2<i1，将(i1-1,j)作为估计LII上的一个新的点，否则，将(i1,j)作为新的估计点；将最新估计点的竖直坐标位置设为i1；(3) If i2 >i1 , take (i1 +1,j) as a new point on estimated LII, if i2 <i1 , take (i1 -1,j) as estimated point on LII A new point, otherwise, take (i1 , j) as a new estimated point; set the vertical coordinate position of the latest estimated point as i1 ;
(4)若j＝N，输出估计的LII；否则，返回步骤(2)。(4) If j=N, output estimated LII; otherwise, return to step (2).
动态规划调整内中膜边界具体步骤是，根据梯度与曲率两方面的因素来建立以下方程：The specific steps of dynamic programming to adjust the intima-media boundary are to establish the following equation according to the two factors of gradient and curvature:
其中，g(xk)表示像素xk位置的梯度值，c(xk-1,xk,xk+1)表示由(xk-1,xk,xk+1)三个点连成折线的曲率大小，λ为一负常数。理想的边界具有梯度值大且曲率小的特点，因此，需要求解(x1,x2,...,xN)使得L(x1,x2,...,xN)函数值达到最大，使用八邻域折现法，将折线上的每个点往上或者往下移动一定像素距离，使得L(x1,x2,...,xN)达到最大值，即得到最合理的边界。Among them, g(xk ) represents the gradient value of the pixel xk position, c(xk-1 ,xk ,xk+1 ) represents the three points from (xk-1 ,xk ,xk+1 ) The curvature of the connected broken line, λ is a negative constant. The ideal boundary has the characteristics of large gradient value and small curvature. Therefore, it is necessary to solve (x1 ,x2 ,...,xN ) so that the function value of L(x1 ,x2 ,...,xN ) reaches Maximum, using the eight-neighborhood discount method, move each point on the broken line up or down by a certain pixel distance, so that L(x1 ,x2 ,...,xN ) reaches the maximum value, that is, the most Reasonable boundaries.
水平集动态规划颈总动脉内中膜厚度测量装置，由超声图像获取装置和计算机构成，计算机对接收到的超声图像后进行处理，计算机上设置有如下模块：The level set dynamic programming common carotid artery intima-media thickness measurement device is composed of an ultrasound image acquisition device and a computer. The computer processes the received ultrasound images. The computer is provided with the following modules:
1)图像裁剪模块：裁剪移除超声图像中非测试信息部分，只保留血管部分；1) Image cropping module: crop and remove the non-test information part in the ultrasound image, and only keep the blood vessel part;
2)提取ROI模块：自动提取ROI；2) Extract ROI module: automatically extract ROI;
3)图像动态拉伸模块：将图像灰度范围统一拉伸至0-255；3) Image dynamic stretching module: uniformly stretch the gray scale range of the image to 0-255;
4)图像滤波模块：采用双边滤波算法去除噪声，同时保留图像边缘信息；4) Image filtering module: use bilateral filtering algorithm to remove noise while retaining image edge information;
5)估计LII模块：采用形态学操作避免内膜缺陷的干扰，使得估计的LII更加靠近真是边界；5) Estimate LII module: use morphological operations to avoid the interference of intimal defects, so that the estimated LII is closer to the real boundary;
6)估计MAI模块：根据梯度图信息估计MAI；6) Estimate MAI module: estimate MAI according to gradient map information;
7)动态规划调整估计边界模块：以梯度值与曲率值的加权和作为动态规划的能量函数，进行迭代计算获得膜边界特性；7) Dynamic programming adjustment estimation boundary module: use the weighted sum of gradient value and curvature value as the energy function of dynamic programming, and perform iterative calculation to obtain membrane boundary characteristics;
8)后处理模块：计算并评估IMT，对自动测量的IMT值进行判别，避免测量结果没有实际意义。8) Post-processing module: calculate and evaluate the IMT, and discriminate the automatically measured IMT value, so as to prevent the measurement results from being meaningless.
本发明的特点及有益效果是：Features and beneficial effects of the present invention are:
水平集算法利用速度函数，在图像的空间范围对水平集函数演化,致使水平集函数的第零水平集所描述的闭合曲线能够在目标轮廓线所在的位置停止。水平集算法能很自然的处理界面拓扑变化，易于求解高维问题。本发明引入水平集方法提取ROI，为后续步骤提供初始图像，通过形态学膨胀腐蚀操作，使图像结果更加可靠，结合动态规划算法，以梯度值与曲率值的加权和作为其能量函数，精确地调整了LII和MAI两条边界，使内中膜厚度测量结果更准确可靠。本发明有力的支持了临床中颈动脉超声图像内中膜厚度的测量，为IMT计算机辅助测量技术的进一步优化发展提供了参考，对专家手动测量的方式是很好的补充。The level set algorithm uses the speed function to evolve the level set function in the spatial range of the image, so that the closed curve described by the zeroth level set of the level set function can stop at the position of the target contour. The level set algorithm can deal with interface topology changes naturally, and is easy to solve high-dimensional problems. The invention introduces the level set method to extract ROI, provides the initial image for the subsequent steps, and makes the image result more reliable through the morphological dilation and erosion operation. Combined with the dynamic programming algorithm, the weighted sum of the gradient value and the curvature value is used as its energy function to accurately The two boundaries of LII and MAI are adjusted to make the measurement results of intima-media thickness more accurate and reliable. The present invention strongly supports the measurement of intima-media thickness in clinical carotid ultrasound images, provides a reference for the further optimization and development of IMT computer-aided measurement technology, and is a good supplement to manual measurement by experts.
本发明中采用水平集方法分割出感兴趣区域，结合动态规划算法对最终的边界进行精确调整。在所给的图像库测试中，本发明能有效完成颈总动脉内中膜边界的分割并有效计算IMT值，具有较好的理论和使用价值。In the present invention, the level set method is used to segment the region of interest, and the final boundary is precisely adjusted in combination with a dynamic programming algorithm. In the given image library test, the invention can effectively complete the segmentation of the intima-media boundary of the common carotid artery and effectively calculate the IMT value, and has good theoretical and application value.
附图说明：Description of drawings:
图1算法流程图；Figure 1 algorithm flow chart;
图2初始颈动脉超声图像；Figure 2 Initial carotid ultrasound image;
图3裁剪并提取ROI的图像；Figure 3 crops and extracts the image of the ROI;
图4动态拉伸的图像结果；Figure 4 The image result of dynamic stretching;
图5滤波前图像；Image before filtering in Figure 5;
图6滤波后图像；Figure 6 Filtered image;
图7初始LII边界；Figure 7 Initial LII boundary;
图8初始LII、MAI边界；Figure 8 initial LII, MAI boundaries;
图9动态规划调整后最终的LII、MAI边界。Fig. 9 The final LII and MAI boundaries after dynamic programming adjustment.
具体实施方式detailed description
本发明采用如下的技术方案：The present invention adopts following technical scheme:
1)图像裁剪。初始超声图像的周围分布有病人和超声仪器的相关信息，这些信息的存在会影响后续的图像处理步骤，因此需要裁剪移除这些部分，只保留血管部分。1) Image cropping. Information about the patient and the ultrasound instrument is distributed around the initial ultrasound image. The existence of this information will affect the subsequent image processing steps. Therefore, it is necessary to crop and remove these parts, and only keep the blood vessel part.
2)ROI提取。自动提取ROI可以避免手动分割ROI的繁琐，实现自动测量IMT的目的。2) ROI extraction. Automatic extraction of ROI can avoid the tedious manual segmentation of ROI and realize the purpose of automatic measurement of IMT.
3)图像动态拉伸。将图像灰度范围统一拉伸至0-255，使图像轮廓更清晰。3) Image dynamic stretching. Stretch the grayscale range of the image uniformly to 0-255 to make the outline of the image clearer.
4)图像滤波。超声图像以实时性、可重复性、非侵入性及成本低的优点，成为颈动脉检查的首选成像方式。双边滤波算法在去除噪声的同时能够较好地保留图像边缘信息。4) Image filtering. With the advantages of real-time, repeatability, non-invasiveness and low cost, ultrasound images have become the preferred imaging method for carotid artery examination. The bilateral filtering algorithm can better preserve image edge information while removing noise.
5)估计LII。形态学操作避免了内膜缺陷的干扰，使得估计的LII更加靠近真是边界。5) Estimate LII. Morphological operations avoid the interference of intimal defects, making the estimated LII closer to the true boundary.
6)估计MAI。根据梯度图信息估计的MAI比单纯的平移估计更加准确。6) Estimate MAI. MAI estimated from gradient map information is more accurate than pure translation estimation.
7)动态规划调整估计边界。以梯度值与曲率值的加权和作为动态规划的能量函数，迭代的结果更加符合颈总动脉血管的膜边界特性。7) Dynamic programming adjusts the estimated boundary. Using the weighted sum of the gradient value and the curvature value as the energy function of dynamic programming, the iterative result is more in line with the membrane boundary characteristics of the common carotid artery.
8)后处理。计算并评估IMT。对自动测量的IMT值进行判别，避免测量结果没有实际意义。8) post-processing. Calculate and evaluate IMT. Discriminate the automatically measured IMT value to prevent the measurement results from being meaningless.
下面结合附图与实例对本发明作进一步说明。The present invention will be further described below in conjunction with accompanying drawings and examples.
1)图像裁剪。在保证远端血管壁存在的条件下，截取超声图像中下方320×300像素大小的区域。1) Image cropping. Under the condition of ensuring the existence of the distal vessel wall, a region with a size of 320×300 pixels in the lower part of the ultrasound image was intercepted.
2)ROI提取。2) ROI extraction.
首先要直接对图像边缘使用连续曲线描述,与此同时采用图像信息对某一能量泛函进行定义，能量泛函的自变量含有边界轮廓曲线。接着采用动态Euler-Lagrange方程的格式得到这个能量泛函相对应的曲线演化的方程；最终采用水平集方法对沿着能量模拟初始曲线下降最快的方向的演化过程进行模拟，以求出最佳的边界轮廓曲线，提取出ROI部分。Firstly, continuous curves should be used to describe the edge of the image directly. At the same time, image information should be used to define an energy functional. The independent variable of the energy functional contains the boundary contour curve. Then use the format of the dynamic Euler-Lagrange equation to obtain the equation of the curve evolution corresponding to the energy functional; finally use the level set method to simulate the evolution process along the direction of the fastest decline of the initial energy simulation curve to find the best The boundary contour curve of the image extracts the ROI part.
3)图像对比度动态拉伸。3) Dynamic stretching of image contrast.
直接灰度变换是采用改变所选图像的灰度范围使图像对比度增强，对比度相对小的图像因此层次更加丰富，由此改善了视觉的感知，最终达到图像增强的效果。将高于m的灰度级进行压缩，直到它压缩到高灰度级的较窄范围内，而将低于m的灰度级进行压缩，使其存在于低灰度级的较窄范围内，拉伸中间的灰度级，可以增强各部分图像的反差，以使图像拥有比较高的对比度。Direct grayscale transformation is to change the grayscale range of the selected image to enhance the contrast of the image. The image with relatively small contrast has richer layers, thereby improving the visual perception and finally achieving the effect of image enhancement. Compress gray levels above m until it is compressed into a narrow range of high gray levels, and compress gray levels below m so that they exist within a narrow range of low gray levels , Stretching the gray level in the middle can enhance the contrast of each part of the image, so that the image has a relatively high contrast.
其中，s表示输出图像的灰度值，r表示输入图像灰度值，E是一个可以控制函数斜率的变量。m作为一个阈值可以被用户根据图像处理方面的实际需要进行参数设置。Among them, s represents the gray value of the output image, r represents the gray value of the input image, and E is a variable that can control the slope of the function. As a threshold value, m can be parameterized by the user according to the actual needs of image processing.
4)双边滤波。4) Bilateral filtering.
对于图像f(x)，基于像素的灰度相似性高斯滤波得到g(x)，数学表达式为：For the image f(x), the Gaussian filter based on the pixel gray similarity obtains g(x), and the mathematical expression is:
其中，为灰度相似性高斯滤波的归一化常数，x,ξ表示像素点，f(x)表示x像素点的灰度值，表示像素ξ与x的灰度相似度。ξ与x的灰度值相差越小，灰度相似度s越大。in, is the normalization constant of the gray similarity Gaussian filter, x, ξ represent the pixel, f(x) represents the gray value of the x pixel, Indicates the gray similarity between pixel ξ and x. The smaller the gray value difference between ξ and x, the larger the gray similarity s.
图像f(x)经过基于像素距离相似性的高斯滤波得到g(x)，数学表达式为：Image f(x) undergoes Gaussian filtering based on pixel distance similarity to obtain g(x), the mathematical expression is:
其中，为距离相似性的高斯滤波的归一化常数，表示像素ξ与x的距离相似度。ξ与x的距离越近，c值越大。in, is the normalization constant of Gaussian filtering for distance similarity, Indicates the distance similarity between pixel ξ and x. The closer the distance between ξ and x, the larger the value of c.
综合灰度相似性和距离相似性这两方面，得到双边滤波公式如下：Combining the two aspects of gray similarity and distance similarity, the bilateral filtering formula is obtained as follows:
其中，为双边滤波的归一化常数，x表示像素点，滤波器为尺寸10×10的矩阵，其中，基于灰度相似性计算出的高斯标准差为σ1＝0.1，基于距离相似性就算出的高斯函数标准差为σ2＝3。in, is the normalization constant of bilateral filtering, x represents a pixel point, and the filter is a matrix with a size of 10×10, where the Gaussian standard deviation calculated based on the gray similarity is σ1 =0.1, and calculated based on the distance similarity The standard deviation of the Gaussian function is σ2 =3.
5)估计LII。5) Estimate LII.
将ROI的灰度值归一化到区间[0,1]范围内，归一化函数的数学表达式如下：Normalize the gray value of the ROI to the interval [0,1], the mathematical expression of the normalization function is as follows:
其中，g(x)为灰度真实值，fmin和fmax分别是ROI内灰度值的最小值和最大值。Among them, g(x) is the true gray value, and fmin and fmax are the minimum and maximum gray values in the ROI, respectively.
归一化后得到的直方图由3个峰和2个谷组成。假设2个谷最低点的灰度值为T1和T2，在ROI中沿竖直方向从上往下搜索，具体搜索步骤如下：The resulting histogram after normalization consists of 3 peaks and 2 valleys. Assuming that the gray values of the lowest points of the two valleys are T1 and T2 , search from top to bottom in the vertical direction in the ROI. The specific search steps are as follows:
(1)令j＝1；将i1从1增加到M进行搜索直到f(i1,j)≥T1；(i1,j)即所估计的LII起点；(1) Set j=1; increase i1 from 1 to M to search until f(i1 ,j)≥T1 ; (i1 ,j) is the estimated starting point of LII;
(2)将j增加为j+1，即在ROI的下一列搜索直到f(i2,j)≥T1。(2) Increase j to j+1, that is, search in the next column of the ROI until f(i2 ,j)≥T1 .
(3)若i2>i1，将(i1+1,j)作为估计LII上的一个新的点，若i2<i1，将(i1-1,j)作为估计LII上的一个新的点，否则，将(i1,j)作为新的估计点；将最新估计点的竖直坐标位置设为i1。(3) If i2 >i1 , take (i1 +1,j) as a new point on estimated LII, if i2 <i1 , take (i1 -1,j) as estimated point on LII A new point, otherwise, take (i1 , j) as a new estimated point; set the vertical coordinate position of the latest estimated point as i1 .
(4)若j＝N，输出估计的LII；否则，返回步骤(2)。(4) If j=N, output estimated LII; otherwise, return to step (2).
6)估计MAI。6) Estimate MAI.
对LII进行形态学膨胀腐蚀后，若将估计的LII往下移动17个像素距离得到粗略的MAI估计。After morphological expansion and erosion of LII, if the estimated LII is moved down by 17 pixels, a rough MAI estimate can be obtained.
7)动态规划调整内中膜边界。7) Dynamic programming adjusts the intima-media boundary.
根据梯度与曲率两方面的因素来建立以下方程：According to the factors of gradient and curvature, the following equations are established:
其中，g(xk)表示像素xk位置的梯度值，c(xk-1,xk,xk+1)表示由(xk-1,xk,xk+1)三个点连成折线的曲率大小，λ为一负常数。理想的边界具有梯度值大且曲率小的特点，因此，需要求解(x1,x2,...,xN)使得L(x1,x2,...,xN)函数值达到最大，使用八邻域折现法，将折线上的每个点往上或者往下移动一定像素距离，使得L(x1,x2,...,xN)达到最大值，即得到最合理的边界。Among them, g(xk ) represents the gradient value of the pixel xk position, c(xk-1 ,xk ,xk+1 ) represents the three points from (xk-1 ,xk ,xk+1 ) The curvature of the connected broken line, λ is a negative constant. The ideal boundary has the characteristics of large gradient value and small curvature. Therefore, it is necessary to solve (x1 ,x2 ,...,xN ) so that the function value of L(x1 ,x2 ,...,xN ) reaches Maximum, using the eight-neighborhood discount method, move each point on the broken line up or down by a certain pixel distance, so that L(x1 ,x2 ,...,xN ) reaches the maximum value, that is, the most Reasonable boundaries.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610782555.5ACN106570856A (en) | 2016-08-31 | 2016-08-31 | Common carotid artery intima-media thickness measuring device and method combining level set segmentation and dynamic programming |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610782555.5ACN106570856A (en) | 2016-08-31 | 2016-08-31 | Common carotid artery intima-media thickness measuring device and method combining level set segmentation and dynamic programming |
| Publication Number | Publication Date |
|---|---|
| CN106570856Atrue CN106570856A (en) | 2017-04-19 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610782555.5APendingCN106570856A (en) | 2016-08-31 | 2016-08-31 | Common carotid artery intima-media thickness measuring device and method combining level set segmentation and dynamic programming |
| Country | Link |
|---|---|
| CN (1) | CN106570856A (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108961278A (en)* | 2018-06-20 | 2018-12-07 | 深圳市旭东数字医学影像技术有限公司 | The method and its system of abdominal wall muscle segmentation based on image data |
| CN109493383A (en)* | 2018-11-23 | 2019-03-19 | 深圳市威尔德医疗电子有限公司 | The measurement method of Internal-media thickness, server and storage medium in ultrasound image |
| CN109615595A (en)* | 2018-12-03 | 2019-04-12 | 中国石油大学(华东) | A Level Set SAR Oil Spill Extraction Method Based on Bilateral Filtering |
| CN109919953A (en)* | 2019-01-21 | 2019-06-21 | 深圳蓝韵医学影像有限公司 | Method, system and the equipment of carotid intimal medial thickness measurement |
| CN110517263A (en)* | 2019-09-02 | 2019-11-29 | 青岛海信医疗设备股份有限公司 | Method, device and storage medium for determining intima-media thickness |
| CN112330634A (en)* | 2020-11-05 | 2021-02-05 | 恒信东方文化股份有限公司 | Method and system for fine edge matting of clothing |
| CN113974688A (en)* | 2021-09-18 | 2022-01-28 | 深圳迈瑞生物医疗电子股份有限公司 | Ultrasonic imaging method and ultrasonic imaging system |
| CN119180814A (en)* | 2024-11-22 | 2024-12-24 | 广东省农业科学院动物科学研究所 | M-type ultrasound-based embrittled tilapia identification method and system |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102446357A (en)* | 2011-11-23 | 2012-05-09 | 浙江工商大学 | Level set SAR (Synthetic Aperture Radar) image segmentation method based on self-adaptive finite element |
| CN102800089A (en)* | 2012-06-28 | 2012-11-28 | 华中科技大学 | Main carotid artery blood vessel extraction and thickness measuring method based on neck ultrasound images |
| CN103093473A (en)* | 2013-01-25 | 2013-05-08 | 北京理工大学 | Multi-target picture segmentation based on level set |
| CN103593848A (en)* | 2013-11-25 | 2014-02-19 | 深圳市恩普电子技术有限公司 | Ultrasonic endocardium tracing method |
| CN103679728A (en)* | 2013-12-16 | 2014-03-26 | 中国科学院电子学研究所 | Water area automatic segmentation method of SAR image of complicated terrain and device |
| CN104665872A (en)* | 2014-12-29 | 2015-06-03 | 深圳开立生物医疗科技股份有限公司 | Ultrasonic image-based carotid intima-media thickness measuring method and device |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102446357A (en)* | 2011-11-23 | 2012-05-09 | 浙江工商大学 | Level set SAR (Synthetic Aperture Radar) image segmentation method based on self-adaptive finite element |
| CN102800089A (en)* | 2012-06-28 | 2012-11-28 | 华中科技大学 | Main carotid artery blood vessel extraction and thickness measuring method based on neck ultrasound images |
| CN103093473A (en)* | 2013-01-25 | 2013-05-08 | 北京理工大学 | Multi-target picture segmentation based on level set |
| CN103593848A (en)* | 2013-11-25 | 2014-02-19 | 深圳市恩普电子技术有限公司 | Ultrasonic endocardium tracing method |
| CN103679728A (en)* | 2013-12-16 | 2014-03-26 | 中国科学院电子学研究所 | Water area automatic segmentation method of SAR image of complicated terrain and device |
| CN104665872A (en)* | 2014-12-29 | 2015-06-03 | 深圳开立生物医疗科技股份有限公司 | Ultrasonic image-based carotid intima-media thickness measuring method and device |
| Title |
|---|
| QIANG LI,WEI ZHANG,ETC.: ""An Improved Approach for Accurate and Efficient Measurement of Common Carotid Artery Intima-Media Thickness in Ultrasound Images"", 《BIOMED RESEARCH INTERNATIONAL》* |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108961278A (en)* | 2018-06-20 | 2018-12-07 | 深圳市旭东数字医学影像技术有限公司 | The method and its system of abdominal wall muscle segmentation based on image data |
| CN108961278B (en)* | 2018-06-20 | 2023-01-31 | 深圳市旭东数字医学影像技术有限公司 | Method and system for abdominal wall muscle segmentation based on image data |
| CN109493383B (en)* | 2018-11-23 | 2022-02-11 | 深圳市威尔德医疗电子有限公司 | Method for measuring intima-media thickness in ultrasonic image, server and storage medium |
| CN109493383A (en)* | 2018-11-23 | 2019-03-19 | 深圳市威尔德医疗电子有限公司 | The measurement method of Internal-media thickness, server and storage medium in ultrasound image |
| CN109615595A (en)* | 2018-12-03 | 2019-04-12 | 中国石油大学(华东) | A Level Set SAR Oil Spill Extraction Method Based on Bilateral Filtering |
| CN109615595B (en)* | 2018-12-03 | 2019-07-23 | 中国石油大学(华东) | A kind of level set SAR oil spilling extracting method based on bilateral filtering |
| CN109919953A (en)* | 2019-01-21 | 2019-06-21 | 深圳蓝韵医学影像有限公司 | Method, system and the equipment of carotid intimal medial thickness measurement |
| CN110517263A (en)* | 2019-09-02 | 2019-11-29 | 青岛海信医疗设备股份有限公司 | Method, device and storage medium for determining intima-media thickness |
| CN112330634A (en)* | 2020-11-05 | 2021-02-05 | 恒信东方文化股份有限公司 | Method and system for fine edge matting of clothing |
| CN113974688A (en)* | 2021-09-18 | 2022-01-28 | 深圳迈瑞生物医疗电子股份有限公司 | Ultrasonic imaging method and ultrasonic imaging system |
| CN113974688B (en)* | 2021-09-18 | 2024-04-16 | 深圳迈瑞生物医疗电子股份有限公司 | Ultrasonic imaging method and ultrasonic imaging system |
| CN119180814A (en)* | 2024-11-22 | 2024-12-24 | 广东省农业科学院动物科学研究所 | M-type ultrasound-based embrittled tilapia identification method and system |
| CN119180814B (en)* | 2024-11-22 | 2025-02-28 | 广东省农业科学院动物科学研究所 | A method and system for identifying brittle tilapia based on M-mode ultrasound |
| Publication | Publication Date | Title |
|---|---|---|
| CN106570856A (en) | Common carotid artery intima-media thickness measuring device and method combining level set segmentation and dynamic programming | |
| CN110136157B (en) | A deep learning-based method for vessel wall segmentation in 3D carotid ultrasound images | |
| CN104794708B (en) | A kind of atherosclerotic plaque ingredient dividing method based on multiple features study | |
| CN113781640A (en) | A method for establishing a 3D face reconstruction model based on weakly supervised learning and its application | |
| CN112446862B (en) | Dynamic breast ultrasound video full-focus real-time detection and segmentation device and system based on artificial intelligence and image processing method | |
| CN102068281B (en) | Processing method for space-occupying lesion ultrasonic images | |
| CN107403201A (en) | Tumour radiotherapy target area and jeopardize that organ is intelligent, automation delineation method | |
| CN108492300B (en) | Pulmonary Vascular Tree Segmentation Method Combined with Tubular Structure Enhancement and Energy Function | |
| CN108053417A (en) | A kind of lung segmenting device of the 3DU-Net networks based on mixing coarse segmentation feature | |
| CN104598891B (en) | Method and device for personal feature analysis based on shoe footprint images | |
| CN113935976B (en) | A method and system for automatic segmentation of blood vessels in organs of enhanced CT images | |
| CN102982547B (en) | Automatic initialized local active contour model heart and cerebral vessel segmentation method | |
| CN102567734B (en) | Specific value based retina thin blood vessel segmentation method | |
| CN105160660A (en) | Active contour blood vessel extraction method and system based on multi-feature Gaussian fitting | |
| CN107680110A (en) | Inner ear three-dimensional level-set segmentation methods based on statistical shape model | |
| CN113782184B (en) | A stroke-assisted assessment system based on pre-learning of facial key points and features | |
| CN107392928A (en) | Breast cancer diagnosis system and method based on MRF Snake segmentations | |
| CN109003280A (en) | Inner membrance dividing method in a kind of blood vessel of binary channels intravascular ultrasound image | |
| CN103455999A (en) | Automatic vessel wall edge detection method based on intravascular ultrasound image sequence | |
| CN110033448A (en) | A kind of male bald Hamilton classification prediction analysis method of AI auxiliary of AGA clinical image | |
| CN104751457B (en) | A kind of new liver segmentation method based on variation energy | |
| Lin et al. | Method for carotid artery 3-D ultrasound image segmentation based on cswin transformer | |
| CN115359002B (en) | A system and method for automatically detecting plaques in carotid artery ultrasound images | |
| CN111339828B (en) | Vein imaging identification method based on the combination of infrared image and ultrasound Doppler | |
| Karthikeyan et al. | Lungs segmentation using multi-level thresholding in CT images |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | Application publication date:20170419 | |
| WD01 | Invention patent application deemed withdrawn after publication |