3D医学图像中标志点的自动识别与定位  

3D Automatic Fiducial Marker Detection and Localization Approach

作　　者：孙曦[1] 周正东[1] 

机构地区：[1]南京航空航天大学医学物理系,南京211100

出　　处：《现代生物医学进展》2008年第4期709-712,共4页Progress in Modern Biomedicine

摘　　要：目的:探索建立一种新的自动识别标志点的方法。方法:本方法主要分成三步:首先,根据标志点的灰度特征在3D图像上搜索标志点,并得到候选点;然后,计算出搜索到的候选点区域的亮度重心,并作为该点的位置坐标;最后,根据标志点大小、相互间位置关系以及标志点周围区域像素的灰度变化等特征,筛选出真正的标志点。结果:利用该算法对三维理想模型和真实CT重建模型上的标志点进行识别,实验的结果表明该算法能准确识别出这两种模型上的标志点,平均误差均小于2个象素。结论:该方法能快速准确地识别出3D医学图像中的标志点。它不需要人为干预,且不受标志点形状的影响。Objective： To explore a new automatic method for 3D detection. Methods： The method included three steps. First, the candidate markers in 3D data set were searched and obtained according to the intensity of fiducial marker. Secondary, the intensity-weighted centroid of each candidate fiducial region was calculated as the candidate markers＇ position. Finally, the real markers were screened out from the candidates by other confined conditions, such as the size of markers, the distance between the markers and the intensity of the region around the marker. Results： The approach was validated by both simulated datasets and CT phantom scan, the average fiducial localization error was less than 2 pixeks. Conclusion： The method provided high 3D localization accuracy and was independent of the shape of the marker.

关 键 词：医学图像标志点 自动识别 目标定位 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]