Information Fusion Using Evidence Theory for Segmentation of Medical Images

摘要：

There are many applications of segmentation in medical imaging. However, the main difficulties are the presence of noise, the large amount of data to be processed and the diversity of imaging modalities. The main interest of evidence theory is that it allows to manage uncertainty and imprecise data, and to perform information fusion coming from neighbor voxels1 to take into account contextual information.We developed a segmentation method based on the evidence theory to face all these difficulties. The calculation of the membership degree of each voxel to each region of interest is computed using basic belief assignment. An information fusion is performed to aggregate information from the voxel to be classified and the neighbor voxels.After presentation of the theoretical aspect of the segmentation strategy, we study the options choose to fix the parameters. Segmentation results on Computed Tomography images are presented. We compare also our algorithm with two algorithms based on the Markov Random Field theory.

关键词： Image segmentation evidence theory information fusion medical imaging

作者: Peng Zhang Isabelle Gardin Patrick Vannorenberghe

作者单位: Laboratoire dInformatique, de Traitement de lInformation et des Systèmes,EA 4108, Université de Ro Laboratoire de Télédétection à Haute Résolution,Université Paul Sabatier Toulouse 3, 31077 TOULOUSE,

会议类型: 国际会议

会议名称: The International Colloquium on Onformation Fusion 2007(2007年国际信息融合研讨会)

会议地点: 西安

会议语种:英文

页码: 283-289

在线出版日期: 2007-08-22（万方平台首次上网日期，不代表论文的发表时间）