Multilinear Subspace Method Based on Geodesic Distance for Volumetric Object Classification

Multilinear Subspace Method Based on Geodesic Distance for Volumetric Object Classification
复制标题

基于测地距离的多线性子空间体物体分类方法

DOI:
10.1007/978-3-030-29888-3_55
复制
发表时间:
2019
期刊:
Lecture Notes in Computer Science
影响因子:
--
通讯作者:
Atsushi Imiya
Atsushi Imiya
中科院分区:
--
文献类型:
--
作者:
Hayato Itoh;Atsushi Imiya

文献摘要

相似文献

器官、器官中的细胞和细胞中的微结构在数学上都是空间纹理。张量允许我们在医学图像计算中直接分析、操作和识别这样的体积数据。基于张量的数据表达利用张量子空间之间的测地线距离为时空体序列的时间形态发生提供了一种分类方法。介绍了多线性空间中Grassmann流形和Stiefel流形的测地线测度。对17例患者的心脏MRI数据集进行了实验评估,结果表明该方法鉴别和分类是有效的。
Organs, cells in organs and microstructures in cells are mathematically spatial textures. Tensors allow us to directly analyse, manipulate and recognise such volumetric data in medical image computing. Tensor-based data expression provides a classification method for temporal morphogenesis of spatiotemporal volumetric sequences using geodesic distances between tensor subspaces. Geodesic measures are introduced both for the Grassmann and Stiefel manifolds in multilinear space. Experimental evaluations of cardiac MRI dataset for 17 patients show the validity of the method for discrimination and classification.