Objective To analyze the diffusion tensor images efficiently with the high dimension of DTI dataset.Methods A novel visualization approach of DTI based on local preserving dimensionality reduction was proposed.The distance between tensors was defined firstly,and the weighted graph was constructed.Then the low dimension data could be calculated based on LPP algorithm.At last,the low dimension data were mapped to RGB color images.R esult Experiments indicated that the algorithm extracted the characteristics of a high-dimensional diffusion tensor image effectively.Conclusion The algorithm can properly capture statistical relationships among tensor image data.%目的:为了更高效地分析弥散张量图像.方法:提出了基于局部保持映射的弥散张量图像降维算法,首先定义了张量之间的距离计算准则,然后根据局部保持映射算法构建张量之间无向连接图,确定张量之间的相似度矩阵,对原始张量数据进行降维,最后将降维获得的数据映射到彩色图像进行显示.结果:通过合成实验和真实实验的数据可以发现,该算法有效提取了高维弥散张量图像的特征.结论:该算法能学习隐藏在高维弥散张量图像中的流形结构,发现张量数据间的内在关系.
展开▼