運用小範圍頻譜之資訊做血管切割之研究

Abstract

由於醫學影像(CTA、MRI、MRA)被廣泛的使用在醫學的診斷上。而許多疾病，像糖尿病、高血壓、動脈硬化等等，都與人體內的管狀結構息息相關。因此若能將醫學影像裡有關於管狀結構之資訊，像彎曲度、管狀物寬度，擷取出來的話，則能輔助醫師作出準確的診斷與了解病情。而在我們的研究裡，提出了一個運用小範圍頻譜的資訊，去做血管切割的演算法。先將某個有管狀結構的ROI做傅立葉轉換，之後去分析頻譜並擷取出管狀結構的方向，根據此方向去找出下一個管狀結構，以迭代的方式逐步的擷取出整條管狀結構。

CTA and MRA are widely used for the diagnosis of serious circulation diseases。Most diseases (ex：diabetes、hypertension and arteriosclerosis) are closely bound up the tube structure of human。This is the reason why segmentation is used for diagnosis。In this paper，we present a method for segmentation of tube structure。Our method use frequency of ROI (Region of Interest) and analysis this information to segment direction of tube structure，and then，according to the direction to find next tube structure，repeatedly to segment entire tube structure。