CN103793711A - Multidimensional vein extracting method based on brain nuclear magnetic resonance image - Google Patents
Multidimensional vein extracting method based on brain nuclear magnetic resonance image Download PDFInfo
- Publication number
- CN103793711A CN103793711A CN201410023305.4A CN201410023305A CN103793711A CN 103793711 A CN103793711 A CN 103793711A CN 201410023305 A CN201410023305 A CN 201410023305A CN 103793711 A CN103793711 A CN 103793711A
- Authority
- CN
- China
- Prior art keywords
- entropy
- energy
- region
- poor
- crowd
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 53
- 210000004556 brain Anatomy 0.000 title claims abstract description 27
- 238000005481 NMR spectroscopy Methods 0.000 title claims abstract 5
- 210000003462 vein Anatomy 0.000 title 1
- 210000001353 entorhinal cortex Anatomy 0.000 claims abstract description 6
- 210000001320 hippocampus Anatomy 0.000 claims abstract description 6
- 208000010877 cognitive disease Diseases 0.000 claims abstract 3
- 208000027061 mild cognitive impairment Diseases 0.000 claims abstract 3
- 208000024827 Alzheimer disease Diseases 0.000 claims description 21
- 230000011218 segmentation Effects 0.000 claims description 7
- 230000032683 aging Effects 0.000 claims 2
- 238000000605 extraction Methods 0.000 abstract description 8
- 230000009466 transformation Effects 0.000 description 18
- 230000006870 function Effects 0.000 description 7
- 230000000694 effects Effects 0.000 description 6
- 238000012706 support-vector machine Methods 0.000 description 6
- 238000003745 diagnosis Methods 0.000 description 5
- 238000012549 training Methods 0.000 description 4
- 238000000354 decomposition reaction Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 230000035945 sensitivity Effects 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000003902 lesion Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 206010003694 Atrophy Diseases 0.000 description 1
- 206010012289 Dementia Diseases 0.000 description 1
- 206010058467 Lung neoplasm malignant Diseases 0.000 description 1
- 230000037444 atrophy Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 238000013399 early diagnosis Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000000971 hippocampal effect Effects 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 210000004072 lung Anatomy 0.000 description 1
- 201000005202 lung cancer Diseases 0.000 description 1
- 208000020816 lung neoplasm Diseases 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- SYHGEUNFJIGTRX-UHFFFAOYSA-N methylenedioxypyrovalerone Chemical compound C=1C=C2OCOC2=CC=1C(=O)C(CCC)N1CCCC1 SYHGEUNFJIGTRX-UHFFFAOYSA-N 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
Images
Classifications
-
- 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
- G06V10/52—Scale-space analysis, e.g. wavelet 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
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
本发明公开了一种基于人群脑部核磁共振图像的多维度纹理提取方法,运用区域增长法将人群脑部核磁共振图像中感兴趣区域分割出来,采用Curvelet变换和Contourlet变换方法提取感兴趣区域的纹理特征参数,其中所述人群包括阿尔茨海默病人群体,轻度认知障碍病人群体和正常老年人群体,所述的感兴趣区域的纹理特征参数包括熵、灰度均值、相关性、能量、同质度、方差、最大概率、逆差距、聚类趋势、对比度、和的均值、差的均值、和的熵、差的熵,所述感兴趣区域包括内嗅皮层和海马两个区域。
The invention discloses a multi-dimensional texture extraction method based on crowd brain nuclear magnetic resonance images, using the region growing method to segment the regions of interest in the crowd brain nuclear magnetic resonance images, and using Curvelet transform and Contourlet transform methods to extract the regions of interest Texture feature parameters, wherein the crowd includes Alzheimer's patient population, mild cognitive impairment patient population and normal elderly population, and the texture feature parameters of the region of interest include entropy, gray mean value, correlation, energy , homogeneity, variance, maximum probability, inverse gap, clustering trend, contrast, mean of sum, mean of difference, entropy of sum, entropy of difference, and the regions of interest include the entorhinal cortex and the hippocampus.
Description
技术领域:Technical field:
本发明属于医学技术领域,具体涉及一种基于脑部核磁共振图像(MRI)的多维度纹理提取方法。The invention belongs to the field of medical technology, in particular to a method for extracting multidimensional textures based on brain magnetic resonance images (MRI).
背景技术:Background technique:
在辅助诊断早期阿尔茨海默病(AD)中,识别MRI图像中ROIs(包括内嗅皮层,海马)的性质具有重要意义。但MRI影像技术只能以海马萎缩作为区别患者和正常人的指标之一,医生对MRI图像的解释易受主观个人影响,缺乏一致性,且不易准确评价痴呆患者症状的严重程度。Identifying the nature of ROIs (including entorhinal cortex, hippocampus) in MRI images is of great significance in aiding the diagnosis of early Alzheimer's disease (AD). However, MRI imaging technology can only use hippocampal atrophy as one of the indicators to distinguish patients from normal people. Doctors' interpretation of MRI images is easily influenced by subjective individuals, lacks consistency, and is difficult to accurately evaluate the severity of symptoms in dementia patients.
1、现有的图像处理技术有以下5种:1. There are five existing image processing technologies:
1)区域增长法(Region-growing Method):1) Region-growing Method:
该方法利用了图像的局部空间信息,可有效克服其它方法存在的图像分割空间不连续的缺点,但尚没有人用该方法进行脑部MRI图像的处理。This method utilizes the local spatial information of the image and can effectively overcome the shortcomings of image segmentation space discontinuity in other methods, but no one has used this method to process brain MRI images.
2)灰度共生矩阵(GLCM):2) Gray level co-occurrence matrix (GLCM):
以往仅有少量相关研究中采用灰度共生矩阵法提取的纹理特征参数,这对于根据脑部MRI图像纹理特征诊断早期AD、MCI还远远不够。In the past, only a small number of relevant studies used the texture feature parameters extracted by the gray-level co-occurrence matrix method, which is far from enough for the diagnosis of early AD and MCI based on the texture features of brain MRI images.
3)小波变换(Wavelet Transformation):3) Wavelet Transformation:
小波变换形成的特征向量虽然能够在一定程度上对图像进行精确描述,但是利用小波变换提取图像中ROIs纹理特征存在着检索精度不高的缺点。Although the feature vector formed by wavelet transform can accurately describe the image to a certain extent, there is a shortcoming of low retrieval accuracy in extracting texture features of ROIs in images using wavelet transform.
4)第二代小波变换(Curvelet变换):4) The second generation wavelet transform (Curvelet transform):
继上世纪80年代后期发展起来的小波变换之后,1996年Swendens提出了先进的第二代小波变换,在基函数算法上也在不断改进,1998年E.J.Candes提出了Ridgelet变换、1999年E.J.Candes和D.L.Donoho发明了Curvelet变换新算法:Following the wavelet transform developed in the late 1980s, Swendens proposed the advanced second-generation wavelet transform in 1996, and the basis function algorithm was also continuously improved. In 1998, E.J.Candes proposed the Ridgelet transform, and in 1999, E.J.Candes and D.L.Donoho invented a new algorithm for Curvelet transformation:
其中:2-j为尺度、θl为方向角θl、为位置R为转换弧度,2006年又提出了快速离散Curvelet变换。第二代小波变换不但保留了小波变换(Wavelet Transformation)方法多尺度的优点,同时还具有各向异性特点,能够很好地逼近奇异曲线,比Wavelet Transformation更加适合分析二维图像中的曲线或边缘特征,而且具有更高的逼近精度和更好的稀疏表达能力,能够为图像提供一种比Wavelet Transformation多分辨分析更加精确的方法。Where: 2 -j is the scale, θ l is the direction angle θ l , for the location R is for converting radians. In 2006, a fast discrete Curvelet transform was proposed. The second-generation wavelet transform not only retains the multi-scale advantages of the wavelet transform (Wavelet Transformation) method, but also has anisotropic characteristics, which can well approximate singular curves, and is more suitable for analyzing curves or edges in two-dimensional images than Wavelet Transformation features, and has higher approximation accuracy and better sparse expression ability, which can provide a more accurate method for images than Wavelet Transformation multi-resolution analysis.
5)Contourlet变换5) Contourlet transformation
Contourlet变换继承了Curvelet变换的各向异性尺度关系,在一定意义上它是Curvelet变换的另一种实现方式。Contourlet变换的基本思想是首先用一个类似小波的多尺度分解捕捉边缘奇异点,再根据方向信息将位置相近的奇异点汇集成轮廓段。Contourlet transform inherits the anisotropic scale relationship of Curvelet transform, and in a certain sense it is another implementation of Curvelet transform. The basic idea of Contourlet transform is to use a wavelet-like multi-scale decomposition to capture edge singular points first, and then gather similar singular points into contour segments according to the direction information.
Contourlet变换可分为两个部分:拉普拉斯塔式滤波器结构(LaplacianPyramid,LP)和二维方向滤波器组(Directional Filter Bank,DFB)。LP分解首先产生原始信号的一个低通采样逼近及原始图像与低通预测图像之间的一个差值图像,对得到的低通图像继续分解得到下一层的低通图像和差值图像,如此逐步滤波得到图像的多分辨率分解;DFB滤波器组使用扇形结构的共轭镜像滤波器组以避免对输入信号的调制,同时将1层二叉树状结构的方向滤波器变成了21个并行通道的结构。Contourlet transform can be divided into two parts: Laplacian Pyramid (LP) and two-dimensional directional filter bank (Directional Filter Bank, DFB). LP decomposition first produces a low-pass sampling approximation of the original signal and a difference image between the original image and the low-pass predicted image, and then continues to decompose the obtained low-pass image to obtain the low-pass image and difference image of the next layer, so The multi-resolution decomposition of the image is obtained by step-by-step filtering; the DFB filter bank uses a fan-shaped conjugate mirror filter bank to avoid modulation of the input signal, and at the same time changes the direction filter of the 1-layer binary tree structure into 21 parallel channels Structure.
Contourlet变换是一种新的图像二维表示方法,具有多分辨率、局部定位、多方向性、近邻界采样和各向异性等性质,其基函数分布于多尺度、多方向上,少量系数即可有效地捕捉图像中的边缘轮廓,而边缘轮廓正是图像中的主要特征。Contourlet transform is a new two-dimensional image representation method, which has the properties of multi-resolution, local positioning, multi-direction, nearest neighbor sampling and anisotropy. Its basis functions are distributed on multi-scale and multi-direction, and a small number of coefficients It can effectively capture the edge contour in the image, and the edge contour is the main feature in the image.
但是这些新方法在处理不同部位的MRI图像时,需要利用基函数重新构造新算法、选取适宜的参数,因此仍有许多理论问题值得研究。Contourlet变换已经成功地用于图像融合等实际问题,而用于脑部图像纹理特征提取的文献报道凤毛麟角。目前所查阅的文献中,仅有人使用GLCM和Wavelet变换对AD组及正常组脑部MRI图像提取纹理特征并建立预测模型,以诊断结果准确性对上述两种方法进行比较,发现采用GLCM提取纹理建模预测效果优于Wavelet变换。目前尚未见有人使用第二代小波变换以及Contourlet变换进行AD脑部MRI图像纹理提取。因此,需要结合上述现有技术的优点,并克服其不足,对图像的处理方法进行改进,以达到提高早期AD、MCI诊断率的目的。However, these new methods need to use basis functions to reconstruct new algorithms and select appropriate parameters when processing MRI images of different parts, so there are still many theoretical issues worthy of research. Contourlet transform has been successfully used in practical problems such as image fusion, but there are few literature reports on brain image texture feature extraction. In the currently reviewed literature, only people use GLCM and Wavelet transform to extract texture features from brain MRI images of AD group and normal group and establish a prediction model. The above two methods are compared with the accuracy of diagnosis results, and it is found that GLCM is used to extract texture features. Modeling prediction effect is better than Wavelet transform. So far, no one has used the second-generation wavelet transform and Contourlet transform to extract the texture of AD brain MRI images. Therefore, it is necessary to combine the advantages of the above-mentioned prior art and overcome its shortcomings to improve the image processing method to achieve the purpose of improving the early diagnosis rate of AD and MCI.
2、常用的预测模型:2. Commonly used forecasting models:
支持向量机(Support Vector Machine,SVM):Support Vector Machine (Support Vector Machine, SVM):
支持向量机是建立在统计学习理论VC维理论和结构风险最小化原理基础上的机器学习方法。其机理是寻找一个满足分类要求的最优分类超平面,使得该超平面在保证分类精度的同时,能够使超平面两侧的空白区域最大化。Support vector machine is a machine learning method based on the VC dimension theory of statistical learning theory and the principle of structural risk minimization. Its mechanism is to find an optimal classification hyperplane that meets the classification requirements, so that the hyperplane can maximize the blank areas on both sides of the hyperplane while ensuring the classification accuracy.
理论上,支持向量机能够实现对线性可分数据的最优分类。以两类数据分类为例,给定训练样本集(xi,yj),i=1,2,…,l,x∈{±1},超平面记作(w·x)+b=0,为使分类面对所有样本正确分类并且具备分类间隔,就要求它满足如下约束:yi[(w·xi)+b]≥1,i=1,2,…,l,可以计算出分类间隔为2/||w||,因此构造最优超平面的问题就转化为在约束式下求:In theory, support vector machines can achieve optimal classification of linearly separable data. Taking two types of data classification as an example, given a training sample set ( xi ,y j ), i=1,2,…,l,x∈{±1}, the hyperplane is denoted as (w·x)+b= 0, in order to make the classification face all samples correctly classified and have a classification interval, it is required to meet the following constraints: y i [(w x i )+b]≥1, i=1,2,...,l, can be calculated The classification interval is 2/||w||, so the problem of constructing the optimal hyperplane is transformed into finding under the constraints:
为了解决该个约束最优化问题,引入Lagrange函数:In order to solve this constrained optimization problem, the Lagrange function is introduced:
式中,αi>0为Lagrange乘数。约束最优化问题的解由Lagrange函数的鞍点决定,并且最优化问题的解在鞍点处满足对w和b的偏导为0,将该QP问题转化为相应的对偶问题即:In the formula, α i >0 is the Lagrange multiplier. The solution of the constrained optimization problem is determined by the saddle point of the Lagrange function, and the solution of the optimization problem satisfies the partial derivative of w and b at the saddle point to be 0, and the QP problem is transformed into the corresponding dual problem:
解得最优解
计算最优权值向量w*和最优偏置b*,分别为:Calculate the optimal weight vector w * and the optimal bias b * , respectively:
式中,下标因此得到最优分类超平面In the formula, the subscript Therefore, the optimal classification hyperplane
(w*·x)+b*=0,而最优分类函数为:(w * x)+b * =0, and the optimal classification function is:
对于线性不可分情况,SVM的主要思想是将输人向量映射到一个高维的特征向量空间,并在该特征空间中构造最优分类面。For the case of linear inseparability, the main idea of SVM is to map the input vector to a high-dimensional feature vector space, and construct the optimal classification surface in this feature space.
将x做从输入空间Rn到特征空间H的变换φ,得:Transform x from the input space R n to the feature space H, get:
x→φ(x)=(φ1(x),φ2(x),…φl(x))T x→φ(x)=(φ 1 (x),φ 2 (x),…φ l (x)) T
以特征向量φ(x)代替输入向量x,则可以得到最优分类函数为:Using the feature vector φ(x) instead of the input vector x, the optimal classification function can be obtained as:
发明内容:Invention content:
本发明的目的在于提供一种对含有MCI、早期AD患者病灶以及正常老年人群ROIs的MRI图像的分割方法以及提取纹理特征的方法,建立多种预测模型,以便更有效的发现MCI患者,诊断早期AD以及观测正常老年人群的脑部结构改变。The purpose of the present invention is to provide a method for segmenting MRI images containing MCI, early AD patient lesions, and normal elderly population ROIs and a method for extracting texture features, and to establish multiple prediction models so as to more effectively find MCI patients and diagnose early AD and observation of brain structural changes in the normal elderly population.
本发明综合了现有技术中的区域增长法、灰度共生矩阵、小波变换等方法的优点,并加以改进,运用Curvelet变换和Contourlet变换提取ROIs的边缘纹理特征,纹理提取方法全面、新颖,能够达到发明目的。The present invention combines the advantages of the region growing method, gray-level co-occurrence matrix, wavelet transform and other methods in the prior art, and improves them, and uses Curvelet transform and Contourlet transform to extract the edge texture features of ROIs. The texture extraction method is comprehensive and novel, and can achieve the purpose of the invention.
本发明用以下主要技术路线建立了一种对含有MCI、早期AD患者病灶以及正常老年人ROIs的MRI图像的分割以及提取纹理特征的方法(具体程序见图4),并据此建立了判断相关ROIs性质的预测模型:The present invention uses the following main technical route to establish a method for segmenting and extracting texture features of MRI images containing MCI, early AD patient lesions, and normal elderly ROIs (see Figure 4 for the specific procedure), and establishes a judgment correlation Predictive models for the properties of ROIs:
1)建立ROIs图像库;1) Establish ROIs image library;
2)运用区域增长法将图像中相关ROIs分割出来;2) Use the region growing method to segment the relevant ROIs in the image;
3)采用Curvelet变换和Contourlet变换处理图像,提取以下各变量:熵(熵)、灰度均值(灰度均值)、相关性(相关性)、能量(能量)、同质度(同质度geneity)、方差(Variance)、最大概率(Maximum probability,最大概率)、逆差距(Inverse Difference Moment,逆差距)、聚类趋势(Cluster Tendency)、对比度(Contrast)、和的均值(Sun-灰度均值)、差的均值(Difference-灰度均值)、和的熵(Sum-熵)、差的熵(Difference-熵);3) Use Curvelet transform and Contourlet transform to process the image, and extract the following variables: entropy (entropy), gray mean (gray mean), correlation (correlation), energy (energy), homogeneity (homogeneity ), variance (Variance), maximum probability (Maximum probability, maximum probability), inverse gap (Inverse Difference Moment, reverse gap), clustering trend (Cluster Tendency), contrast (Contrast), and the mean (Sun-gray mean ), the mean of the difference (Difference-gray mean), the entropy of the sum (Sum-entropy), and the entropy of the difference (Difference-entropy);
4)用步骤2)~3)所得到的各种变量数据建立图像特征参量数据库;4) Use the various variable data obtained in steps 2) to 3) to establish an image characteristic parameter database;
5)根据步骤4)的数据库构建基于Curvelet变换和Contourlet变换的预测模型;建立预测模型的方法包括支持向量机;5) According to the database in step 4), a prediction model based on Curvelet transformation and Contourlet transformation is constructed; the method for establishing the prediction model includes a support vector machine;
6)将步骤5)所得到的各种参量数据与样本经反复验证,以修正预测模型,得到结果比较准确的理想模型。6) The various parameter data and samples obtained in step 5) are repeatedly verified to correct the prediction model and obtain an ideal model with relatively accurate results.
7)比较步骤5)中基于Curvelet变换和Contourlet变换建立预测模型的预测效果。7) Compare the prediction effect of the prediction model based on Curvelet transformation and Contourlet transformation in step 5).
目前对于图像的纹理还没有一个统一的定义,一般认为图像的纹理特征描述的是物体表面灰度或颜色的变化,这种变化与物体的自身属性有关,是某种纹理基元的重复。At present, there is no unified definition of the texture of an image. It is generally believed that the texture feature of an image describes the change of grayscale or color on the surface of an object. This change is related to the object's own attributes and is the repetition of a certain texture primitive.
采用Curvelet变换和Contourlet变换可以得到以下纹理特征参数:The following texture feature parameters can be obtained by using Curvelet transformation and Contourlet transformation:
熵熵:反映影像纹理的随机性;entropy entropy: Reflect the randomness of image texture;
灰度均值灰度均值:反映像素所有灰度值的集中趋势。Grayscale mean Grayscale mean: Reflects the central tendency of all gray values of pixels.
相关性相关性:测量像素灰度的相关性;Correlation Correlation: Measure the correlation of pixel gray levels;
能量(Angular Second Moment)能量(角二阶矩):反映影像灰度分布的均匀程度和纹理粗细度;Energy (Angular Second Moment) Energy (angular second moment): Reflect the uniformity of image grayscale distribution and texture thickness;
同质度geneity同质度:反映灰度水平相似程度;Homogeneity degree of geneity degree of homogeneity: Reflect the similarity of the gray level;
Variance方差:反映灰度水平的分布情况;Variance variance: Reflect the distribution of gray levels;
Maximum probability(最大概率)最大概率:最突出的像素对的发生率;Maximum probability (maximum probability) maximum probability: The incidence of the most salient pixel pair;
Inverse Difference Moment(逆差距)逆差矩:反映图像的平滑性;Inverse Difference Moment (inverse gap) inverse difference moment: Reflect the smoothness of the image;
Cluster tendency聚类趋势:测量相似灰度水平值像素的分组;Cluster tendency clustering trend: Measuring groupings of pixels with similar gray level values;
Contrast对比度:反映影像的清晰度;Contrast contrast: reflect the clarity of the image;
Sun-灰度均值和的均值Difference-灰度均值差的均值提供图像中灰度水平的均值。Sun - the mean of the sum of the gray mean Difference - the mean of the gray mean difference Provides the mean of the gray levels in the image.
Sum-熵和的熵Difference-熵差的熵
本发明用上述方法建立了早期AD预测模型,判断准确度达到了100%。The present invention uses the above method to establish an early AD prediction model, and the judgment accuracy reaches 100%.
以下是ROIs内部纹理提取的实例,步骤如下:The following is an example of texture extraction inside ROIs, the steps are as follows:
1、收集AD、MCI及正常老年人的MRI原始图像(附图说明以正常老年人的MRI图像为例),见图1;1. Collect the original MRI images of AD, MCI and normal elderly people (the illustration of the figure takes MRI images of normal elderly people as an example), see Figure 1;
2、用区域增长法分割上述图像,得到图像见图2、图3,分割采用编写好的程序,直接运行即可。2. Use the region growth method to segment the above image, and the obtained images are shown in Fig. 2 and Fig. 3. The segmented program is written and can be run directly.
3、采用Curvelet变换和Contourlet变换提取纹理特征参量,每种方法分别有相应程序,直接运行即可,提取的纹理特征参量见表2~表29。3. Use Curvelet transform and Contourlet transform to extract texture feature parameters. Each method has a corresponding program, which can be run directly. The extracted texture feature parameters are shown in Table 2 to Table 29.
实验结果证明:Experimental results prove that:
表1Curvelet变换与Contourlet变化不同方位差异性分析Table 1 Analysis of the difference between Curvelet transformation and Contourlet transformation in different orientations
由表1可知,Contourlet变换比Curvelet变换能较好反映AD组,MCI组,正常组之间纹理值总体差异性。It can be seen from Table 1 that the Contourlet transformation can better reflect the overall difference in texture value among the AD group, the MCI group, and the normal group than the Curvelet transformation.
采用区域增长法分割ROIs,选取80%的数据作为训练样本,再根据余下20%的数据作为验证样本,以检验通过模型判断ROIs是否病变与病理诊断的一致性,证明通过Contourlet提取脑部MRI图像ROIs外部纹理建立预测早期肺癌的灵敏度、特异度均为100%。Using the region growth method to segment ROIs, select 80% of the data as training samples, and then use the remaining 20% of the data as verification samples to test whether the ROIs are determined by the model. The sensitivity and specificity of ROIs external texture to predict early lung cancer were both 100%.
通过以上数据,可以得到结论:采用区域增长法分割脑部MRI图像,并通Contourlet变换提取纹理特征参量建立预测模型对早期AD辅助诊断有很好的效果。Based on the above data, it can be concluded that using the region growing method to segment brain MRI images, and extracting texture feature parameters through Contourlet transform to establish a predictive model has a good effect on the early auxiliary diagnosis of AD.
有益效果:Beneficial effect:
将区域增长法用于脑部MRI的分割,这是本发明的一个创新。经实验证明,区域增长法分割提取纹理建模要更优于整体分割提取纹理建模,能更好的保留结节的边缘信息;It is an innovation of the present invention that the region growing method is used for the segmentation of brain MRI. Experiments have proved that the segmentation and extraction texture modeling of the region growing method is better than the overall segmentation and extraction texture modeling, and can better preserve the edge information of nodules;
Contourlet是提取脑部MRI图像内部纹理特征的有效方法,可以提取14种纹理特征参数,较全面的反映图像的纹理特征。Contourlet is an effective method for extracting internal texture features of brain MRI images. It can extract 14 kinds of texture feature parameters and reflect the texture features of images more comprehensively.
附图说明:Description of drawings:
图1是正常老年人的MRI原始图像,其中用黑框标注的是ROIs;Figure 1 is the original MRI image of a normal elderly person, where ROIs are marked with black boxes;
图2是图1中用区域增长法分割后得到的左脑ROIs放大图像;Fig. 2 is the magnified image of the left brain ROIs obtained after segmentation by the region growing method in Fig. 1;
图3是图1中用区域增长法分割后得到的右脑ROIs放大图像;Figure 3 is an enlarged image of the right brain ROIs obtained after segmentation by the region growing method in Figure 1;
图4是用本发明方法对脑部MRI图像进行多维度纹理提取的技术路线;Fig. 4 is the technical route of carrying out multi-dimensional texture extraction to brain MRI image with the method of the present invention;
其中,1是左脑海马、内嗅皮层区域;2是右脑海马、内嗅皮层区域。Among them, 1 is the left hippocampus and entorhinal cortex area; 2 is the right hippocampus and entorhinal cortex area.
具体实施方式Detailed ways
以下实例是用本发明方法建立基于含相关ROIs的脑部MRI图像预测AD的模型介绍,这只是对本发明方法的进一步说明,但实例并不限制本发明的应用范围。实际上,用本方法还可对其它类型的医学图像进行性质判断。The following example is a model introduction based on the method of the present invention to establish AD prediction based on brain MRI images containing relevant ROIs. This is only a further description of the method of the present invention, but the examples do not limit the scope of application of the present invention. In fact, this method can also be used to judge the properties of other types of medical images.
图像来源:ADNI网站上共享的AD、MCI及正常老年人的脑部MRI图像,分别为.Nii格式的图像,使用MRIcro软件读取;Image source: Brain MRI images of AD, MCI and normal elderly shared on the ADNI website, respectively in .Nii format, read by MRIcro software;
方法:采用Matlab软件编程,运用区域增长法将上述MRI图像中的ROIs分割出来,利用Contourlet变换提取相关ROIs的纹理特征参数。Methods: Matlab software was used to program, the ROIs in the above MRI images were segmented by region growing method, and the texture feature parameters of relevant ROIs were extracted by Contourlet transform.
以下是脑部MRI图像ROIs纹理特征参量提取的实例,步骤如下:The following is an example of extracting texture feature parameters of brain MRI image ROIs, the steps are as follows:
1、分别收集AD组病例20例,MCI病例组20例,正常组病例20例总共的250张脑部MRI原始图像,病例年龄为40~89岁,平均年龄为64岁,年龄分布的中位数为60岁。1. A total of 250 original brain MRI images were collected from 20 cases of AD group, 20 cases of MCI case group, and 20 cases of normal group. The age of the cases ranged from 40 to 89 years old, the average age was 64 years old, and the median age distribution was The number is 60 years old.
2、用区域增长法分割图像,分割方式按区域增长法常规方法,采用编写好的程序,设定阈值为35,直接运行。图1是待分割的一个图像举例,得到图像见图2、图3,2. Segment the image with the region growing method, the segmentation method follows the conventional method of the region growing method, use the written program, set the threshold to 35, and run it directly. Figure 1 is an example of an image to be segmented, see Figure 2 and Figure 3 for the obtained image,
3、分别采用Curvelet变换和Contourlet变换提取纹理特征参量,提取的纹理特征参量见表2~表29。3. Using Curvelet transform and Contourlet transform to extract texture feature parameters respectively, the extracted texture feature parameters are shown in Table 2-Table 29.
4、采用Curvelet变换和Contourlet变换方法提取脑部MRI图像ROIs分别得到14个纹理特征参量建立预测模型。4. Use Curvelet transform and Contourlet transform methods to extract brain MRI image ROIs to obtain 14 texture feature parameters to establish prediction models.
5、比较Curvelet变换和Contourlet变换两种纹理提取方法建立预测模型的预测效果。5. Compare the prediction effect of two texture extraction methods, Curvelet transform and Contourlet transform, to establish a predictive model.
由12例200张脑部MRI图像作为训练样本集,余下的3例50张脑部MRI图像作为测试样本进行支持向量机进行分类预测,预测效果:灵敏度=100%;特异度=100%;符合率=100%。The 200 brain MRI images of 12 cases were used as the training sample set, and the remaining 3 cases of 50 brain MRI images were used as test samples for classification prediction by support vector machine. The prediction effect: sensitivity = 100%; specificity = 100%; Rate = 100%.
实验结果证明:采用区域增长法分割ROIs,选取80%的数据作为训练样本,再根据余下的20%的数据作为验证样本,以检验通过模型判断肺ROIs是否病变与病理诊断的一致性,证明通过对小样本脑部MRI图像ROIs提取纹理特征参量预测早期AD的灵敏度为100%。The experimental results prove that: using the region growth method to segment ROIs, select 80% of the data as training samples, and then use the remaining 20% of the data as verification samples to test whether the lung ROIs are determined by the model. The sensitivity of extracting texture feature parameters from small sample brain MRI image ROIs to predict early AD is 100%.
通过以上数据,可以得到结论:采用区域增长法分割脑部MRI图像ROIs,并通过Contourlet变换提取纹理特征参量建立预测模型对早期AD的辅助诊断有很好的效果。Based on the above data, it can be concluded that using the region growing method to segment brain MRI image ROIs, and extracting texture feature parameters through Contourlet transform to establish a prediction model has a good effect on the auxiliary diagnosis of early AD.
Curvelet变换提取的变量见表2~表15,Contourlet变换提取的变量见表16~表29(以AD组编号为136_s_0194的病例纹理值为例)。The variables extracted by Curvelet transformation are shown in Tables 2 to 15, and the variables extracted by Contourlet transformation are shown in Tables 16 to 29 (taking the texture value of the case number 136_s_0194 in the AD group as an example).
表2灰度均值Table 2 gray mean value
表3标准差Table 3 standard deviation
表4聚类趋势Table 4 Clustering trend
表5同质度Table 5 Homogeneity
表6最大概率Table 6 Maximum probability
表7能量Table 7 Energy
表8惯性矩Table 8 Moments of inertia
表9逆差距Table 9 Inverse Gap
表10熵Table 10 Entropy
表11相关性Table 11 Correlation
表12和的均数Table 12 and mean of
表13差的均数Table 13 Mean of difference
表14和的熵Table 14 and the entropy of
表15差的熵Table 15 Entropy of difference
表16灰度均值Table 16 gray mean value
表17标准差Table 17 Standard Deviation
表18聚类趋势Table 18 Clustering trend
表19同质度Table 19 Homogeneity
表20最大概率Table 20 Maximum Probability
表21能量Table 21 Energy
表22惯性矩Table 22 Moments of inertia
表23逆差距Table 23 Inverse Gap
表24熵熵Table 24 Entropy Entropy
表25相关性Table 25 Correlation
表26和的均数Table 26 and the mean of
表27差的均数Table 27 Mean of difference
表28和的熵Entropy of Table 28 and
表29差的熵Table 29 Entropy of difference
Claims (3)
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410023305.4A CN103793711A (en) | 2014-01-17 | 2014-01-17 | Multidimensional vein extracting method based on brain nuclear magnetic resonance image |
PCT/CN2014/001130 WO2015106374A1 (en) | 2014-01-17 | 2014-12-16 | Multidimensional texture extraction method based on brain nuclear magnetic resonance images |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410023305.4A CN103793711A (en) | 2014-01-17 | 2014-01-17 | Multidimensional vein extracting method based on brain nuclear magnetic resonance image |
Publications (1)
Publication Number | Publication Date |
---|---|
CN103793711A true CN103793711A (en) | 2014-05-14 |
Family
ID=50669354
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410023305.4A Pending CN103793711A (en) | 2014-01-17 | 2014-01-17 | Multidimensional vein extracting method based on brain nuclear magnetic resonance image |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN103793711A (en) |
WO (1) | WO2015106374A1 (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2015106374A1 (en) * | 2014-01-17 | 2015-07-23 | 首都医科大学 | Multidimensional texture extraction method based on brain nuclear magnetic resonance images |
WO2015106373A1 (en) * | 2014-01-17 | 2015-07-23 | 首都医科大学 | Method for establishing prediction model based on multidimensional texture of brain nuclear magnetic resonance images |
CN104881680A (en) * | 2015-05-25 | 2015-09-02 | 电子科技大学 | Alzheimer's disease and mild cognitive impairment identification method based on two-dimension features and three-dimension features |
CN106780372A (en) * | 2016-11-30 | 2017-05-31 | 华南理工大学 | A kind of weight nuclear norm magnetic resonance imaging method for reconstructing sparse based on Generalized Tree |
CN109492700A (en) * | 2018-11-21 | 2019-03-19 | 西安中科光电精密工程有限公司 | A kind of Target under Complicated Background recognition methods based on multidimensional information fusion |
CN110033019A (en) * | 2019-03-06 | 2019-07-19 | 腾讯科技(深圳)有限公司 | Method for detecting abnormality, device and the storage medium of human body |
CN111739645A (en) * | 2020-05-14 | 2020-10-02 | 上海依智医疗技术有限公司 | Training method of immune-related pneumonia prediction model |
CN111754598A (en) * | 2020-06-27 | 2020-10-09 | 昆明理工大学 | Transform Learning-Based Parallel MRI Reconstruction Method for Local Spatial Neighborhood |
CN113962930A (en) * | 2021-09-07 | 2022-01-21 | 北京邮电大学 | Alzheimer's disease risk assessment model building method and electronic device |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108241865B (en) * | 2016-12-26 | 2021-11-02 | 哈尔滨工业大学 | Multi-level quantitative staging method for liver fibrosis based on multi-scale and multi-subgraphs of ultrasound images |
US20210228079A1 (en) | 2018-03-07 | 2021-07-29 | Institut National De La Sante Et De La Recherche Medicale (Inserm) | Method for early prediction of neurodegenerative decline |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101971213A (en) * | 2008-02-29 | 2011-02-09 | 新加坡科技研究局 | A method and system for anatomy structure segmentation and modeling in an image |
CN101853493B (en) * | 2009-10-21 | 2013-06-19 | 首都医科大学 | Method for extracting multi-dimensional texture of nodi from medical images |
CN103793908A (en) * | 2014-01-17 | 2014-05-14 | 首都医科大学 | Method for constructing prediction model of multifunctional veins based on brain nuclear magnetic resonance image |
CN103793711A (en) * | 2014-01-17 | 2014-05-14 | 首都医科大学 | Multidimensional vein extracting method based on brain nuclear magnetic resonance image |
-
2014
- 2014-01-17 CN CN201410023305.4A patent/CN103793711A/en active Pending
- 2014-12-16 WO PCT/CN2014/001130 patent/WO2015106374A1/en active Application Filing
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2015106373A1 (en) * | 2014-01-17 | 2015-07-23 | 首都医科大学 | Method for establishing prediction model based on multidimensional texture of brain nuclear magnetic resonance images |
WO2015106374A1 (en) * | 2014-01-17 | 2015-07-23 | 首都医科大学 | Multidimensional texture extraction method based on brain nuclear magnetic resonance images |
CN104881680A (en) * | 2015-05-25 | 2015-09-02 | 电子科技大学 | Alzheimer's disease and mild cognitive impairment identification method based on two-dimension features and three-dimension features |
CN106780372A (en) * | 2016-11-30 | 2017-05-31 | 华南理工大学 | A kind of weight nuclear norm magnetic resonance imaging method for reconstructing sparse based on Generalized Tree |
CN106780372B (en) * | 2016-11-30 | 2019-06-18 | 华南理工大学 | A Weighted Kernel Norm Magnetic Resonance Imaging Reconstruction Method Based on Generalized Tree Sparse |
CN109492700B (en) * | 2018-11-21 | 2020-09-08 | 西安中科光电精密工程有限公司 | Complex background target identification method based on multi-dimensional information fusion |
CN109492700A (en) * | 2018-11-21 | 2019-03-19 | 西安中科光电精密工程有限公司 | A kind of Target under Complicated Background recognition methods based on multidimensional information fusion |
CN110033019A (en) * | 2019-03-06 | 2019-07-19 | 腾讯科技(深圳)有限公司 | Method for detecting abnormality, device and the storage medium of human body |
CN110033019B (en) * | 2019-03-06 | 2021-07-27 | 腾讯科技(深圳)有限公司 | Method and device for detecting abnormality of human body part and storage medium |
CN111739645A (en) * | 2020-05-14 | 2020-10-02 | 上海依智医疗技术有限公司 | Training method of immune-related pneumonia prediction model |
CN111739645B (en) * | 2020-05-14 | 2024-01-30 | 北京深睿博联科技有限责任公司 | Training method of immune-related pneumonia prediction model |
CN111754598A (en) * | 2020-06-27 | 2020-10-09 | 昆明理工大学 | Transform Learning-Based Parallel MRI Reconstruction Method for Local Spatial Neighborhood |
CN111754598B (en) * | 2020-06-27 | 2022-02-25 | 昆明理工大学 | Transform Learning-Based Parallel MRI Reconstruction Method for Local Spatial Neighborhood |
CN113962930A (en) * | 2021-09-07 | 2022-01-21 | 北京邮电大学 | Alzheimer's disease risk assessment model building method and electronic device |
CN113962930B (en) * | 2021-09-07 | 2022-09-09 | 北京邮电大学 | Alzheimer disease risk assessment model establishing method and electronic equipment |
Also Published As
Publication number | Publication date |
---|---|
WO2015106374A1 (en) | 2015-07-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103793711A (en) | Multidimensional vein extracting method based on brain nuclear magnetic resonance image | |
Zhang et al. | Automatic skin lesion segmentation by coupling deep fully convolutional networks and shallow network with textons | |
CN113902761B (en) | Knowledge distillation-based unsupervised segmentation method for lung disease focus | |
CN103793908A (en) | Method for constructing prediction model of multifunctional veins based on brain nuclear magnetic resonance image | |
CN108428229A (en) | It is a kind of that apparent and geometric properties lung's Texture Recognitions are extracted based on deep neural network | |
CN104881680A (en) | Alzheimer's disease and mild cognitive impairment identification method based on two-dimension features and three-dimension features | |
CN111462116A (en) | Multimodal parameter model optimization fusion method based on radiomics features | |
CN103400154B (en) | A kind of based on the human motion recognition method having supervision Isometric Maps | |
CN115496771A (en) | A Brain Tumor Segmentation Method Based on Brain 3D MRI Image Design | |
Yang et al. | Skull sex estimation based on wavelet transform and Fourier transform | |
CN108549912A (en) | A kind of medical image pulmonary nodule detection method based on machine learning | |
CN111340770B (en) | A cancer prognostic model building method combining global weighted LBP and texture analysis | |
CN108319969B (en) | A method and system for glioma survival prediction based on a sparse representation framework | |
CN105809175A (en) | Encephaledema segmentation method and system based on support vector machine algorithm | |
CN104268873A (en) | Breast tumor partition method based on nuclear magnetic resonance images | |
CN118262875A (en) | Medical image diagnosis and contrast film reading method | |
CN101609485B (en) | Medical Imaging Diagnosis System and Diagnosis Method Based on Migration Kernel Matching and Tracking | |
CN104281856A (en) | Image preprocessing method and system for brain medical image classification | |
CN101853493B (en) | Method for extracting multi-dimensional texture of nodi from medical images | |
CN105512670A (en) | HRCT peripheral nerve cutting based on KECA feature dimension reduction and clustering | |
CN108090507A (en) | A kind of medical imaging textural characteristics processing method based on integrated approach | |
Roy et al. | Heterogeneity of human brain tumor with lesion identification, localization, and analysis from MRI | |
CN113610746A (en) | Image processing method and device, computer equipment and storage medium | |
CN112614093A (en) | Breast pathology image classification method based on multi-scale space attention network | |
Zeng et al. | Abnormality detection via iterative deformable registration and basis-pursuit decomposition |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20140514 |