Improved reproducibility of neuroanatomical definitions through diffeomorphometry and complexity reduction

Daniel Tward, Jorge Jovicich, Andrea Soricelli, Giovanni Frisoni, Alain Trouvé, Laurent Younes, Michael Miller

Research output: Contribution to journalArticle

Abstract

We present an algorithm for passing from dense noisy neuroanatomical segmentations, directly to a complexity-reduced representation with respect to a deformed smooth template surface, bypassing the need for triangulation of any target data. We demonstrate the utility of this algorithm toward improving reproducibility of hippocampal definitions, using a dataset containing 4 MR images per subject, two within the same visit on each of two dates, with dense segmentations provided by unedited longitudinal Freesurfer analysis. We quantify reproducibility of intra-visit and inter-visit variability through L2 distances and Hausdorff distances between pairs of segmentations, and show that our method results in a statistically significant improvement by a factor of 1.63 to more than 3-fold.

Original languageEnglish
Pages (from-to)223-230
Number of pages8
JournalLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume8679
Publication statusPublished - 2014

Keywords

  • Complexity reduction
  • Diffeomorphometry
  • LDDMM
  • Neuroanatomy
  • Neuroimaging
  • Reproducibility

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Fingerprint Dive into the research topics of 'Improved reproducibility of neuroanatomical definitions through diffeomorphometry and complexity reduction'. Together they form a unique fingerprint.

  • Cite this