WO2011160309A1 - 基于鲁棒统计信息传播的多模态三维磁共振图像脑肿瘤分割方法 - Google Patents

基于鲁棒统计信息传播的多模态三维磁共振图像脑肿瘤分割方法 Download PDF

Info

Publication number
WO2011160309A1
WO2011160309A1 PCT/CN2010/074517 CN2010074517W WO2011160309A1 WO 2011160309 A1 WO2011160309 A1 WO 2011160309A1 CN 2010074517 W CN2010074517 W CN 2010074517W WO 2011160309 A1 WO2011160309 A1 WO 2011160309A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
magnetic resonance
information
multimodal
segmentation
Prior art date
Application number
PCT/CN2010/074517
Other languages
English (en)
French (fr)
Inventor
范勇
李宏明
Original Assignee
中国科学院自动化研究所
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 中国科学院自动化研究所 filed Critical 中国科学院自动化研究所
Priority to CN201080001752XA priority Critical patent/CN102596025B/zh
Priority to PCT/CN2010/074517 priority patent/WO2011160309A1/zh
Priority to US13/000,255 priority patent/US9129382B2/en
Publication of WO2011160309A1 publication Critical patent/WO2011160309A1/zh

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/10Segmentation; Edge detection
    • G06T7/11Region-based segmentation
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/10Segmentation; Edge detection
    • G06T7/143Segmentation; Edge detection involving probabilistic approaches, e.g. Markov random field [MRF] modelling
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/10Segmentation; Edge detection
    • G06T7/162Segmentation; Edge detection involving graph-based methods
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/10Image acquisition modality
    • G06T2207/10072Tomographic images
    • G06T2207/10088Magnetic resonance imaging [MRI]
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/20Special algorithmic details
    • G06T2207/20081Training; Learning
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/30Subject of image; Context of image processing
    • G06T2207/30004Biomedical image processing
    • G06T2207/30016Brain
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/30Subject of image; Context of image processing
    • G06T2207/30004Biomedical image processing
    • G06T2207/30096Tumor; Lesion

Definitions

  • the invention relates to the technical field of medical image processing, and in particular to a method for automatically segmenting brain tumors on a multimodal magnetic resonance brain image. Background technique
  • the graph cut method divides the image based on the region information and the boundary information of the image.
  • the region information is measured by calculating the probability that the image pixel belongs to a certain region
  • the boundary information is measured by calculating the similarity between the image pixels, including the pixel.
  • the similarity between gray levels and the proximity of spatial locations In the existing work, Wels (Wels et al., "Method and System for Brain Tumor Segmentation in 3D Magneti c Resonance Images", United States Patent Applic Publication, Pub.
  • Brain tumor segmentation is performed based on a PBT (probabi listic boosting tree) classifier and a graph cut method.
  • the PBT classifier trained on the training image set is used to measure the region information for the voxels in the image to be segmented; the boundary information is measured by the gray level similarity between adjacent image voxels.
  • the classification performance of PBT classifier based on supervised learning depends on the consistency of the image gray information in the training image set and the gray information of the image to be segmented. For the magnetic resonance image, due to the noise, the deviation field and the patient during the acquisition process. Differences between individuals, there are often large grayscale differences between individual images.
  • the regional information metrics that are completely dependent on training statistics are not completely reliable, and some unreliable information metrics will affect subsequent graph cuts. result.
  • the measure of boundary information only considers the relationship between directly adjacent voxels, resulting in the final graph cut result being not robust to image noise, even a small amount of noise will have a greater impact on the minimum cut result of the graph, and thus affect The final image segmentation boundary.
  • the purpose of the method is to provide a method for automatic segmentation of brain tumors on multimodal three-dimensional magnetic resonance images.
  • a method for automatically segmenting a brain tumor for a multimodal three-dimensional magnetic resonance image includes the steps of:
  • the tumor region segmentation result is obtained based on the initial tag information and the representation of the graph.
  • the invention realizes the local and global consistency of the gray information of the image to be segmented, thereby realizing the segmentation of the tumor region.
  • This embodiment can reduce the influence of gray scale differences between individual patient images and insufficient training statistical information on the segmentation results to some extent.
  • Figure 1 is a multimodal magnetic resonance brain image containing a tumor
  • Fig. 3 is a result of segmentation of the brain tumor image of Fig. 1. detailed description
  • Multimodal three-dimensional magnetic resonance images include T1-weighted images, T2-weighted images, and contrast-enhanced T1-weighted images.
  • images 102, 104, and 106 respectively show axial view of a layer of different modal magnetic resonance images, where 102 is a T1-weighted image, 104 is a T2-weighted image, and 106 is a contrast-enhanced T1 weighting.
  • Image 108 gives the results of standard segmentation of the tumor area.
  • the method comprehensively utilizes a statistical discriminant method and a graph-based label propagation method for segmentation of a tumor region.
  • the support vector machine-based classifier uses the image context information of each voxel to classify and discriminate, which is used to initialize the segmentation of the tumor region.
  • the graph-based label propagation method is based on the gray similarity between the voxels and the spatial structure proximity. And multi-modal image edge information, the final segmentation result is obtained by optimizing the corresponding objective function.
  • FIG. 2 shows the flow of brain tumor segmentation using multi-modal magnetic resonance images using this method.
  • a multimodal three-dimensional magnetic resonance image is acquired, including a T1-weighted image, a T2-weighted image, and a contrast-enhanced T1-weighted image. Therefore, each voxel in the multimodal image contains three gray values corresponding to the images of the respective pulse sequences.
  • the multi-modal three-dimensional magnetic resonance image is preprocessed, which mainly includes: 1) image registration between multi-modal images to eliminate possible head motion effects; 2) removing non-multi-modal images Brain tissue; 3) Image bias field correction to reduce grayscale inhomogeneity of the image; 4) Grayscale normalization between individual patient images.
  • Specific implementation methods can be found in the literature: Smith et al., "Advances in Functional and Structural MR image Analysis and Implementation as FSL", Neurolmage 23 (2004), pgs. 208 219, Smi th, " Fast robust Automated Brain Extraction", Human Brain Mapping 17 (2002), pgs.
  • the support vector machine based classifier trained on the training image set is used to classify each voxel in the segmented image and calculate its probability of belonging to the tumor region.
  • the classifier based on support vector machine can combine multiple image features and find the optimal hyperplane in the feature space, so that the classification result has the largest classification boundary and the smallest classification error.
  • the method uses the multimodal image gray information of all nodes in its spatial neighborhood to form features.
  • Vector as input to the support vector machine.
  • the relevant parameters used to train the support vector machine are determined by cross-validation.
  • the support vector machine classifies each voxel into a tumor region or a normal tissue and outputs a corresponding classification probability, which indicates the reliability of the classification result.
  • the initial label information of the image segmentation is determined according to the classification probability obtained by the support vector machine: the voxel whose classification probability is higher than a certain threshold is marked as the voxel of the initial mark.
  • a representation of the map is created for the magnetic resonance image to be segmented.
  • the graph-based segmentation method models the image to be segmented into a weighted map GO''. £), where each node in 1'' corresponds to each voxel in the image, and each edge in £ corresponds to a pair of nodes and Weights are given to describe the similarity of the features of the nodes at both ends.
  • the tumor segmentation problem is transformed into a label value that assigns a foreground or background to each node in the graph, representing the tumor region and the normal tissue region, respectively. It is known from the clustering assumption that neighboring nodes or nodes on the same feature structure may have the same tag value. Therefore, given the tag information of the partially marked node, the tag value of the unmarked node can be inferred from the tag information of the tag node based on the feature consistency between the nodes.
  • This markup problem can be obtained by solving the following objective functions:
  • Q(F) F T (J-5 ⁇ F + ii(F - h ⁇ (J ⁇ (1) where / is the unit matrix, 5 is the normalized similarity matrix, and f is the final label information, which is the initial
  • the first item in the formula is a local consistency constraint in which the neighboring node has similar label information
  • the second item is a constraint on the consistency of the final labeling result and the initial label information, and the two are obtained by parameter trade-off. Marked results that satisfy local and global consistency.
  • the edge weights are defined as follows:
  • step 212 based on the initial tag information in step 208 and the map established in step 210, the objective function in equation (1) is optimized to obtain the final tumor region segmentation result.
  • the segmented brain tumor region is output: the segmented brain tumor region can be obtained by overlaying a binary mask image onto a multimodal magnetic resonance image.
  • Figure 3 shows the results obtained by performing tumor segmentation on the magnetic resonance brain image shown in Figure 1 using this method.
  • image 302 is the original contrast-enhanced T1-weighted image (same as 106 in FIG. 1);
  • image 304 is the result of tumor segmentation using a support vector machine-based classifier alone;
  • image 306 is for tumor and normal tissue
  • the result of the segmentation marker initialization red is a tumor, blue is normal tissue);
  • image 308 is the final tumor segmentation result obtained using this method. As can be seen by comparing image 308 with 108 in Figure 1, the method is capable of accurately segmenting brain tumor regions.
  • the present invention utilizes the robust statistical information provided by the support vector machine to guide the graph-based label propagation method for segmentation.
  • the features and innovations mainly lie in: 1) use in the training image set Statistical discriminant model to provide reliable statistical tumor information; 2) Use semi-supervised learning method based on graph theory to propagate reliable statistical tumor information to the entire image space according to the local and global consistency of the image to be segmented; 3) Using robustness
  • the "Boundary Stop" function enhances the description of the edge weights in the graph to limit the information propagation between nodes in different attribute regions of the image.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Probability & Statistics with Applications (AREA)
  • Software Systems (AREA)
  • Magnetic Resonance Imaging Apparatus (AREA)

Description

基于鲁棒统计信息传播的多模态三维磁共振图像脑肿瘤分割方法 技术领域
本发明涉及医学图像处理技术领域,特别涉及对多模态磁共振脑图像 进行脑肿瘤自动分割的方法。 背景技术
在医学影像诊断中,对磁共振脑图像中的肿瘤区域进行检测和分割具 有重要的意义。可靠、精确地分割脑肿瘤区域不仅能够为手术计划的制定 和治疗效果的评估提供有用的信息,而且可以用来对病变组织进行建模以 及构建带有病变的脑图谱为科学研究所利用。然而, 当前在临床活动中往 往是临床专家对肿瘤区域进行手动分割, 这不仅费时、 费力, 而且不同时 间点、 由不同专家分割的结果经常存在较大的差异。 因此, 全自动的脑肿 瘤图像分割算法实施起来具有一定的优势。 尽管研究人员为了开发高效、 准确、可重复的脑肿瘤分割算法已经付出了巨大的努力, 但是由于病人个 体间脑肿瘤的形状、 大小、位置之间存在很大差异, 开发鲁棒的自动分割 算法仍然是个很大的挑战。
近年来, 基于图的图像分割方法越来越多地被应用于医学图像分割, 如图割、随机行走等。 图割的方法基于图像的区域信息和边界信息对图像 进行分割, 区域信息通过计算图像像素隶属于某一区域的概率进行度量, 边界信息通过计算图像像素之间的相似性进行度量,包含像素之间灰度的 相似性以及空间位置的邻近性。 在已有的工作中, Wels (Wels et al. , "Method and System for Brain Tumor Segmentation in 3D Magneti c Resonance Images ", United States Patent Appl ication Publication, Pub. No.: US 2010/0027865 ),提出利用基于 PBT(probabi listic boosting tree)的分类器和图割的方法进行脑肿瘤分割。其中, 使用在训练图像集 上训练得到的 PBT分类器为待分割图像中的体素进行区域信息的度量;边 界信息通过相邻图像体素间的灰度相似性度量。基于有监督学习的 PBT分 类器的分类表现依赖于训练图像集中图像灰度信息和待分割图像灰度信 息的一致性, 对于磁共振图像, 由于采集过程中的噪声、 偏差场以及病人 个体之间的差异, 个体图像之间往往存在较大的灰度差异, 因此, 完全依 赖于训练统计信息得到的区域信息度量并不完全可靠,部分不可靠的信息 度量将会影响后续的图割结果。此外, 边界信息的度量只考虑直接相邻体 素之间的联系, 导致最终的图割结果对于图像噪声不鲁棒, 即使少量的噪 声都会对图的最小割结果产生较大的影响, 进而影响最终的图像分割边 界。 发明内容
本方法的目的是提供一种对多模态三维磁共振图像进行脑肿瘤自动 分割的方法。
为了实现上述目的,一种对多模态三维磁共振图像进行脑肿瘤自动分 割的方法, 包括步骤:
对输入的多模态三维磁共振图像进行预处理;
对预处理后的待分割图像中的每个体素进行分类并计算其隶属于脑 肿瘤区域的概率, 根据分类概率确定图像分割的初始标签信息;
对预处理后的待分割的图像建立图的表示;
根据初始标签信息和图的表示求得肿瘤区域分割结果。
本发明通过发掘待分割图像自身灰度信息局部和全局的一致性,进而 实现肿瘤区域的分割。这种实施方式能够在一定程度上减少病人个体图像 间灰度差异和训练统计信息不充分对分割结果造成的影响。 附图说明
图 1是包含肿瘤的多模态磁共振脑图像;
图 2是本方法一种实施方式的流程图;
图 3是对图 1 中脑肿瘤图像进行分割的结果。 具体实施方式
下面结合附图详细说明本发明技术方案中所涉及的各个细节问题。应 指出的是, 所描述的实施例仅旨在便于对于本发明的理解, 而不起任何限 定作用。 动分割。 多模态三维磁共振图像包括 Tl加权像, T2加权像以及对比度增 强的 T1加权像。 如图 1所示, 图像 102, 104, 106分别展示了一层不同 模态磁共振图像的轴状位视图, 其中, 102 为 T1加权像, 104为 T2加权 像, 106为对比度增强的 T1加权像; 图像 108给出了肿瘤区域标准分割 的结果。
本方法综合利用统计判别方法和基于图的标签传播方法进行肿瘤区 域的分割。基于支持向量机的分类器利用各体素的图像上下文信息进行分 类判别, 进而用来初始化肿瘤区域的分割; 基于图的标签传播方法依据图 像体素之间的灰度相似性、空间结构邻近性以及多模态图像边缘信息, 通 过优化相应的目标函数求得最终的分割结果。
图 2 展示了使用本方法进行多模态磁共振图像脑肿瘤分割的流程。 在步骤 202中, 获取多模态三维磁共振图像, 其中包括 T1加权像, T2加权像, 对比度增强的 T1加权像。 因此, 多模态图像中的每个体素包 含三个灰度值, 对应于各个脉冲序列的图像。
在步骤 204中, 对多模态三维磁共振图像进行预处理, 主要包括: 1) 多模态图像间的图像配准以消除可能存在的头动影响; 2) 去除多模态图 像中的非脑组织; 3) 图像偏差场校正以减轻图像的灰度不均匀性; 4) 不 同病人个体图像间的灰度标准化。具体实施方法可参阅文献: Smith et al . , "Advances in Functional and Structural MR image Analysis and Implementation as FSL", Neurolmage 23 (2004) , pgs. 208 219, Smi th, " Fast robust Automated Brain Extraction ", Human Brain Mapping 17 (2002) , pgs. 143-155, Sled et al. , "A Nonpar ame trie Method for Automatic Correction of Intensi ty Non-uniformity in MR I Data", IEEE Trans. Med. Imaging 17 (1998) , pgs. 87 97。
在步骤 206中,使用在训练图像集上训练得到的基于支持向量机的分 类器对待分割图像中每个体素进行分类并计算其属于肿瘤区域的概率。基 于支持向量机的分类器能够结合多种图像特征,并在特征空间中寻找最优 超平面, 使分类结果具有最大的分类边界以及最小的分类误差。对于每个 体素,本方法使用其空间邻域内所有节点的多模态图像灰度信息构成特征 向量, 作为支持向量机的输入。训练支持向量机所使用的相关参数通过交 叉验证的方法确定。支持向量机将每个体素分类为肿瘤区域或正常组织并 输出相应的分类概率, 这个概率表示了分类结果的可靠性。根据由支持向 量机得到的分类概率确定图像分割的初始标签信息:分类概率高于一定阈 值的体素被标记为初始标记的体素。
在步骤 210中, 为待分割的磁共振图像建立图的表示。基于图的分割 方法将待分割图像建模为一个加权图 GO''. £),其中 1''中的每个节点对应图像 中的每个体素, £中的每条边对应一对节点并且被赋予权重以描述两端节 点特征的相似性。在图的表示下, 肿瘤区域分割问题转化为对图中每个节 点赋予前景或背景的标签值, 分别表示肿瘤区域和正常组织区域。 由聚类 假设可知,邻近的节点或者处于相同特征结构上的节点可能具有相同的标 签值。 因此, 给定部分标记节点的标签信息, 未标记节点的标签值可以根 据节点间的特征一致性由标记节点的标签信息推测获得。此标记问题可以 通过求解下列目标函数获得:
Q(F) = FT(J― 5}F + ii(F - h^ (J― (1) 其中 /为单位矩阵, 5为规整化的相似性矩阵, f为最终标签信息, 为初 始的标签信息。式中的第一项是邻近节点具有相似的标签信息的局部一致 性约束条件, 第二项是最终标记结果和初始标签信息的一致性的约束条 件, 两者通过参数权衡以求得满足局部和全局一致性的标记结果。
由于未标记节点基于边的权重吸收来自其他节点的标签信息,所以部 分初始可靠标记的节点以及定义合理的边权重对于成功地进行图像分割 具有重要影响。 在本方法中, 初始标签信息^ t.由基于支持向量机的分类 器提供。在计算图中边权重的过程中, 本方法全面地考虑了多模态图像灰 度信息的相似性、空间解剖结构的邻近性以及多模态图像的边缘信息, 边 权重定义如下:
其中 和 . 分别度量多模态图像灰度信息的相似性和空间结构邻近 性, 4 是和多模态图像边缘信息有关的数据项, 和 J'为图中不同的节点。 和 计算如下: - h l< r , (4)
Figure imgf000007_0001
rwise 其中 为节点 A的多模态图像灰度向量, k.为节点 的空间位置, 和 控制相应核函数的尺度。多模态图像边缘信息被嵌入到一个 "边界停止" 函数中, 该函数可以为任意对图像噪声鲁棒的单调递减函数。 本实现中, 度量图像边缘信息的数据项计算如下:
!/ G;i ≤ ¾ , , (5) i 0 otherwise
其中 为图像中沿节点 i j方向上 i-J'之间的最大图像梯度幅值, ¾为控 制函数尺度的参数。 此数据项用来指示节点间是否存在图像边缘, 值 比较小意味着节点 和 j处于相同区域的概率比较低, 从而应该限制两 者之间的标签信息传播。 的值可以通过鲁棒统计的技巧来估计。
在步骤 212中,基于在步骤 208中的初始标签信息和在步骤 210中建 立的图, 优化式(1)中的目标函数, 从而求得最终的肿瘤区域分割结果。
Q(F) 的极小化问题可以通过迭代的方式求得, 计算过程如下:
F"t+1 = (1 ~ }SFm + aL,t , (6) 其中 ^为第 A次迭代更新得到的标签信息, F。和 i:fU相同, 0 < 《1是和《相 关的参数, 用来平衡来自其他节点的信息以及节点自身的初始信息。此迭 代过程可以看作标签信息传播的过程: 在每次迭代中, 每个节点吸收来自 其他节点的信息, 并保持部分自身的初始信息。 当迭代过程收敛时, 每 个未标记的节点被标记为其吸收最多标签信息的类别,从而得到相应的二 值掩膜图像。
关于图的建立及目标函数优化过程可参阅文献: Zhou et al.,
"Learning with local and global consistency", Advances in Neural Information Processing Systems (2004) , pgs. 321 328 在步骤 214中, 输出分割得到的脑肿瘤区域: 分割的脑肿瘤区域可以 通过将二值掩膜图像覆盖到多模态磁共振图像上得到。
图 3给出了使用本方法对图 1中所示磁共振脑图像进行肿瘤分割得到 的结果。其中, 图像 302 为原始的对比度增强的 T1加权像(和图 1中 106 相同); 图像 304为单独使用基于支持向量机的分类器进行肿瘤分割得到 的结果; 图像 306为对肿瘤和正常组织进行分割标记初始化的结果 (红色 为肿瘤, 蓝色为正常组织); 图像 308为使用本方法得到的最终肿瘤分割 结果。通过图像 308和图 1中的 108相比较可以看出, 本方法能够准确地 对脑肿瘤区域进行分割。
不同于目前常用的脑肿瘤图像分割方法,本发明利用由支持向量机提 供的鲁棒统计信息指导基于图论的标签传播方法进行分割,其特色和创新 主要在于: 1) 使用在训练图像集中得到的统计判别模型以提供可靠的统 计肿瘤信息; 2) 使用基于图理论的半监督学习方法将可靠的统计肿瘤信 息根据待分割图像局部和全局的一致性传播到整个图像空间; 3) 使用鲁 棒的 "边界停止"函数加强对图中边权重的描述, 以限制图像不同属性区 域节点之间的信息传播。
以上所述, 仅为本发明中的具体实施方式, 但本发明的保护范围并不 局限于此, 任何熟悉该技术的人在本发明所披露的技术范围内, 可理解想 到的变换或替换, 都应涵盖在本发明的范围之内, 因此, 本发明的保护范 围应该以权利要求书的保护范围为准。

Claims

权 利 要 求
1. 一种对多模态三维磁共振图像进行脑肿瘤自动分割的方法, 包括 步骤:
对输入的多模态三维磁共振图像进行预处理;
对预处理后的待分割图像中的每个体素进行分类并计算其隶属于脑 肿瘤区域的概率, 根据分类概率确定图像分割的初始标签信息;
对预处理后的待分割的图像建立图的表示;
根据初始标签信息和图的表示求得肿瘤区域分割结果。
2. 根据权利要求 1所述的方法, 其特征在于使用基于支持向量机的 分类器对待分割的图像中的体素进行分类并计算所述概率。
3. 根据权利要求 1所述的方法, 其特征在于所述初始标签信息包括: 肿瘤分类概率高于一定阈值的体素被标记为基于图的分割方法的初 始标记。
4. 根据权利要求 1所述的方法, 其特征在于对于每一个体素, 根据 空间邻域内所有体素的多模态灰度信息构建其特征向量。
5. 根据权利要求 1所述的方法, 其特征在于所述图的表示包括: 每 个节点对应于多模态磁共振图像中的一个体素, 每条边的权重 对应于所连接两节点的特征相似性度量。
6. 根据权利要求 5所述的方法, 其特征在于图标识中各节点对应一 标签值,根据各节点对应的由基于支持向量机的分类器得到的分类概率判 断其分类是否可靠,并将可靠分类的节点作为基于图的分割方法的初始标 签, 未可靠分类的节点不作处理。
7. 按权利要求 5 所述的方法, 其特征在于图节点之间的相似性度量 包括:
节点之间的相似性度量由多模态图像灰度信息的相似性、结构空间 位置的邻近性以及节点间多模态图像边缘信息三部分组成。
8. 按权利要求 1所述的方法, 其特征在于基于初始标记信息以及图 的边权重使用迭代的方法计算最优解。
PCT/CN2010/074517 2010-06-25 2010-06-25 基于鲁棒统计信息传播的多模态三维磁共振图像脑肿瘤分割方法 WO2011160309A1 (zh)

Priority Applications (3)

Application Number Priority Date Filing Date Title
CN201080001752XA CN102596025B (zh) 2010-06-25 2010-06-25 基于鲁棒统计信息传播的多模态三维磁共振图像脑肿瘤分割方法
PCT/CN2010/074517 WO2011160309A1 (zh) 2010-06-25 2010-06-25 基于鲁棒统计信息传播的多模态三维磁共振图像脑肿瘤分割方法
US13/000,255 US9129382B2 (en) 2010-06-25 2010-06-25 Method and system for brain tumor segmentation in multi-parameter 3D MR images via robust statistic information propagation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2010/074517 WO2011160309A1 (zh) 2010-06-25 2010-06-25 基于鲁棒统计信息传播的多模态三维磁共振图像脑肿瘤分割方法

Publications (1)

Publication Number Publication Date
WO2011160309A1 true WO2011160309A1 (zh) 2011-12-29

Family

ID=45370850

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2010/074517 WO2011160309A1 (zh) 2010-06-25 2010-06-25 基于鲁棒统计信息传播的多模态三维磁共振图像脑肿瘤分割方法

Country Status (3)

Country Link
US (1) US9129382B2 (zh)
CN (1) CN102596025B (zh)
WO (1) WO2011160309A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108538369A (zh) * 2018-03-20 2018-09-14 中南大学湘雅医院 中枢神经系统肿瘤影像数据的分析方法
CN113160142A (zh) * 2021-03-24 2021-07-23 浙江工业大学 一种融合先验边界的脑肿瘤分割方法

Families Citing this family (32)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8908948B2 (en) * 2011-12-21 2014-12-09 Institute Of Automation, Chinese Academy Of Sciences Method for brain tumor segmentation in multi-parametric image based on statistical information and multi-scale structure information
US9135695B2 (en) * 2012-04-04 2015-09-15 Siemens Aktiengesellschaft Method for creating attenuation correction maps for PET image reconstruction
US9405959B2 (en) 2013-03-11 2016-08-02 The United States Of America, As Represented By The Secretary Of The Navy System and method for classification of objects from 3D reconstruction
WO2014152919A1 (en) * 2013-03-14 2014-09-25 Arizona Board Of Regents, A Body Corporate Of The State Of Arizona For And On Behalf Of Arizona State University Kernel sparse models for automated tumor segmentation
US9977107B2 (en) 2013-04-03 2018-05-22 Siemens Healthcare Gmbh Atlas-free brain tissue segmentation method using a single T1-weighted MRI acquisition
US10169685B2 (en) 2014-07-07 2019-01-01 The Regents Of The University Of California Automatic segmentation and quantitative parameterization of brain tumors in MRI
ES2537153B2 (es) * 2014-09-05 2015-10-08 Universitat Politècnica De València Método y sistema de generación de imágenes nosológicas multiparamétricas
DE102014115851A1 (de) * 2014-10-30 2016-05-04 Physikalisch - Technische Bundesanstalt Verfahren und Vorrichtung zur Berechnung, Darstellung und Weiterverarbeitung von lokalen Gütemaßen aus einem Volumenbilddatensatz
RU2713707C2 (ru) * 2015-04-30 2020-02-06 Конинклейке Филипс Н.В. Классификация тканей головного мозга
EP3289563B1 (en) 2015-04-30 2020-08-05 Koninklijke Philips N.V. Brain tissue classification
CN104834943A (zh) * 2015-05-25 2015-08-12 电子科技大学 一种基于深度学习的脑肿瘤分类方法
US10909675B2 (en) 2015-10-09 2021-02-02 Mayo Foundation For Medical Education And Research System and method for tissue characterization based on texture information using multi-parametric MRI
US10325412B2 (en) 2015-11-05 2019-06-18 Shenyang Neusoft Medical Systems Co., Ltd. Cutting three-dimensional image
CN105608690B (zh) * 2015-12-05 2018-06-08 陕西师范大学 一种基于图论和半监督学习相结合的图像分割方法
CN105718962B (zh) * 2016-03-09 2019-03-29 绍兴文理学院 基于图像子块支持向量机的脑血流信号计算方法
CN105931253B (zh) * 2016-05-16 2018-11-06 陕西师范大学 一种基于半监督学习相结合的图像分割方法
WO2018014108A1 (en) * 2016-07-22 2018-01-25 UNIVERSITé LAVAL System and method for estimating synthetic quantitative health values from medical images
US9870614B1 (en) * 2016-07-25 2018-01-16 Sony Corporation Automatic 3D brain tumor segmentation and classification
CN106504245A (zh) * 2016-10-28 2017-03-15 东北大学 一种多模态脑影像损伤性病变组织图像分割方法
US10176570B2 (en) 2016-11-16 2019-01-08 Sony Corporation Inter-patient brain registration
RU2654199C1 (ru) * 2017-07-18 2018-05-16 Самсунг Электроникс Ко., Лтд. Сегментация тканей человека на компьютерном изображении
US11288803B2 (en) * 2017-10-09 2022-03-29 Koninklijke Philips N.V. Ablation result validation system
CN108986115B (zh) * 2018-07-12 2020-12-18 佛山生物图腾科技有限公司 医学图像分割方法、装置及智能终端
CN109784287A (zh) * 2019-01-22 2019-05-21 中国科学院自动化研究所 基于情景信号类前额叶网络的信息处理方法、系统、装置
CN110363103B (zh) * 2019-06-24 2021-08-13 仲恺农业工程学院 虫害识别方法、装置、计算机设备及存储介质
CN111091574B (zh) * 2019-12-21 2024-04-02 中国人民解放军第四军医大学 一种基于单像素特征的医学图像分割方法
CN115461781A (zh) * 2020-02-20 2022-12-09 得克萨斯州大学系统董事会 用于经由大脑解剖学的多模态3d分析优化探针在脑部中的规划和放置的方法
CN112508902B (zh) * 2020-11-30 2024-05-14 上海联影智能医疗科技有限公司 白质高信号分级方法、电子设备及存储介质
CN113379758B (zh) * 2021-05-01 2024-04-12 首都医科大学宣武医院 一种脑图像的迭代分割方法、设备及系统
EP4346674A1 (en) * 2021-06-03 2024-04-10 Case Western Reserve University Systems, methods, and media for presenting biophysical simulations in an interactive mixed reality environment
CN113777546B (zh) * 2021-09-13 2023-04-25 西安交通大学医学院第一附属医院 一种基于三维映射的磁共振多参数脑白质高信号量化方法
CN117437493B (zh) * 2023-12-20 2024-03-29 泰山学院 联合一阶和二阶特征的脑肿瘤mri图像分类方法及系统

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070064983A1 (en) * 2005-09-16 2007-03-22 Wen-Chen Huang Method for automatically detecting nasal tumor
CN101061509A (zh) * 2004-11-19 2007-10-24 皇家飞利浦电子股份有限公司 用于医疗成像数据内的肿瘤边界的自动检测和分割的系统和方法
CN100378752C (zh) * 2005-06-03 2008-04-02 中国科学院自动化研究所 鲁棒的自然图像分割方法
US20080170769A1 (en) * 2006-12-15 2008-07-17 Stefan Assmann Method and image processing system for producing result images of an examination object
CN101334895A (zh) * 2008-08-07 2008-12-31 清华大学 一种针对动态增强乳腺磁共振影像序列的影像分割方法
CN100586368C (zh) * 2006-03-20 2010-02-03 沈阳东软医疗系统有限公司 基于人体解剖结构对称性的磁共振成像扫描自动定位方法

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2007079207A2 (en) * 2005-12-30 2007-07-12 Yeda Research & Development Co. Ltd. An integrated segmentation and classification approach applied to medical applications analysis
US7865002B2 (en) * 2006-06-02 2011-01-04 Carolina Imaging Specialists, Inc. Methods and apparatus for computer automated diagnosis of mammogram images
US7983459B2 (en) * 2006-10-25 2011-07-19 Rcadia Medical Imaging Ltd. Creating a blood vessel tree from imaging data
EP2136331A1 (en) * 2008-06-17 2009-12-23 Siemens Schweiz AG Method for segmentation of an MRI image of a tissue in presence of partial volume effects
US8280133B2 (en) * 2008-08-01 2012-10-02 Siemens Aktiengesellschaft Method and system for brain tumor segmentation in 3D magnetic resonance images

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101061509A (zh) * 2004-11-19 2007-10-24 皇家飞利浦电子股份有限公司 用于医疗成像数据内的肿瘤边界的自动检测和分割的系统和方法
CN100378752C (zh) * 2005-06-03 2008-04-02 中国科学院自动化研究所 鲁棒的自然图像分割方法
US20070064983A1 (en) * 2005-09-16 2007-03-22 Wen-Chen Huang Method for automatically detecting nasal tumor
CN100586368C (zh) * 2006-03-20 2010-02-03 沈阳东软医疗系统有限公司 基于人体解剖结构对称性的磁共振成像扫描自动定位方法
US20080170769A1 (en) * 2006-12-15 2008-07-17 Stefan Assmann Method and image processing system for producing result images of an examination object
CN101334895A (zh) * 2008-08-07 2008-12-31 清华大学 一种针对动态增强乳腺磁共振影像序列的影像分割方法

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108538369A (zh) * 2018-03-20 2018-09-14 中南大学湘雅医院 中枢神经系统肿瘤影像数据的分析方法
CN108538369B (zh) * 2018-03-20 2022-02-15 中南大学湘雅医院 中枢神经系统肿瘤影像数据的分析方法
CN113160142A (zh) * 2021-03-24 2021-07-23 浙江工业大学 一种融合先验边界的脑肿瘤分割方法

Also Published As

Publication number Publication date
US20120093381A1 (en) 2012-04-19
CN102596025A (zh) 2012-07-18
CN102596025B (zh) 2013-10-16
US9129382B2 (en) 2015-09-08

Similar Documents

Publication Publication Date Title
WO2011160309A1 (zh) 基于鲁棒统计信息传播的多模态三维磁共振图像脑肿瘤分割方法
US8908948B2 (en) Method for brain tumor segmentation in multi-parametric image based on statistical information and multi-scale structure information
US11094067B2 (en) Method and system for image processing
WO2021017372A1 (zh) 一种基于生成对抗网络的医学图像分割方法、系统及电子设备
Hamamci et al. Tumor-cut: segmentation of brain tumors on contrast enhanced MR images for radiosurgery applications
Hua et al. Segmenting brain tumor using cascaded V-Nets in multimodal MR images
Roy et al. A deep learning-shape driven level set synergism for pulmonary nodule segmentation
Keraudren et al. Automated fetal brain segmentation from 2D MRI slices for motion correction
Ma et al. A review of algorithms for medical image segmentation and their applications to the female pelvic cavity
Hua et al. Segmentation of pathological and diseased lung tissue in CT images using a graph-search algorithm
Garcia-Lorenzo et al. Trimmed-likelihood estimation for focal lesions and tissue segmentation in multisequence MRI for multiple sclerosis
JP6005297B2 (ja) 磁気共鳴(mr)イメージにおける骨と軟骨の分類方法、および、データ処理システム
Mahapatra Cardiac image segmentation from cine cardiac MRI using graph cuts and shape priors
Karimaghaloo et al. Automatic detection of gadolinium-enhancing multiple sclerosis lesions in brain MRI using conditional random fields
Wen et al. A novel statistical cerebrovascular segmentation algorithm with particle swarm optimization
WO2007044508A2 (en) System and method for whole body landmark detection, segmentation and change quantification in digital images
Menze et al. Segmenting glioma in multi-modal images using a generative model for brain lesion segmentation
JP2014502169A (ja) 医用画像のセグメンテーションのためのシステム
Mahapatra et al. Active learning based segmentation of Crohn's disease using principles of visual saliency
Mahapatra et al. Visual saliency based active learning for prostate mri segmentation
Zeng et al. Liver segmentation in magnetic resonance imaging via mean shape fitting with fully convolutional neural networks
WO2018176189A1 (zh) 图像分割的方法及系统
Li et al. Automated segmentation of prostate MR images using prior knowledge enhanced random walker
Chama et al. Automated lung field segmentation in CT images using mean shift clustering and geometrical features
Hao et al. Local label learning (L3) for multi-atlas based segmentation

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 201080001752.X

Country of ref document: CN

WWE Wipo information: entry into national phase

Ref document number: 13000255

Country of ref document: US

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

Ref document number: 10853452

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: 10853452

Country of ref document: EP

Kind code of ref document: A1