Power Euclidean metrics for covariance matrices with application to diffusion tensor imaging

2010/09/15 by Ian L. Dryden, Xavier Pennec, Dryden, Ian L. +3 · 1 citation
Mathematics · #Applications (stat.AP) #FOS: Computer and information sciences #Methodology (stat.ME) #stat.AP #stat.ME

paper · pdf · doi:10.48550/arxiv.1009.3045

arxiv created 2010/09/15 · arxiv updated 2010/09/17

Abstract

Various metrics for comparing diffusion tensors have been recently proposed in the literature. We consider a broad family of metrics which is indexed by a single power parameter. A likelihood-based procedure is developed for choosing the most appropriate metric from the family for a given dataset at hand. The approach is analogous to using the Box-Cox transformation that is frequently investigated in regression analysis. The methodology is illustrated with a simulation study and an application to a real dataset of diffusion tensor images of canine hearts.

Cited by

• Doubly robust estimation of causal effects for random object outcomes with continuous treatments

Related

• Non-Euclidean statistical analysis of covariance matrices and diffusion tensors
• An anisotropy preserving metric for DTI processing
• Log-Euclidean Signatures for Intrinsic Distances Between Unaligned Datasets
• Statistical tests for group comparison of manifold-valued data
• spd-metrics-id: A Python Package for SPD-Aware Distance Metrics in Connectome Fingerprinting and Beyond