公开(公告)号：US08768024B1

公开(公告)日：2014-07-01

申请号：US13907076

申请日：2013-05-31

Applicant: Given Imaging Ltd.

Inventor： Igor Zingman ,  Ofra Zinaty

IPC: G06K9/00

CPC classification number: G06T5/30 ,  A61B1/00009 ,  A61B1/041 ,  G06T7/44 ,  G06T2207/10068 ,  G06T2207/20021 ,  G06T2207/30028 ,  G06T2207/30092

Abstract: A method and system for automatically detecting villi texture in in-vivo images. In-vivo images may be received from an in vivo imaging capsule, and at least one image may be selected for analysis. Textural details may be extracted, for example using morphological top hat transforms. The image may be divided into segments or patches. The patches may be pruned using a set of threshold conditions, which are calculated based on grayscale or color statistics of the patches. The patches may be classified as containing villi texture or not based for example on rotation-invariant textural features computed in the Fourier domain. A sum, score or rating for the image may be calculated based on the classification of each patch. The final score of the image may be averaged over a number of, for example, consecutive images, and the score may be used for detecting entry into the small bowel and segmenting the image stream to anatomical regions.

Abstract translation: 一种自动检测体内图像绒毛纹理的方法和系统。 可以从体内成像胶囊接收体内图像，并且可以选择至少一个图像用于分析。 可以提取纹理细节，例如使用形态顶帽变换。 图像可以分为片段或片段。 可以使用一组阈值条件修剪补丁，这些阈值条件是基于补丁的灰度或颜色统计计算的。 斑块可以被分类为包含绒毛纹理或者不是基于例如在傅立叶域中计算的旋转不变纹理特征。 可以基于每个补丁的分类来计算图像的总和，得分或评级。 图像的最终分数可以在多个例如连续图像上进行平均，并且分数可以用于检测进入小肠并将图像流分割成解剖区域。