Page 136 - Contributed Paper Session (CPS) - Volume 3

CPS1972 Livio Corain et al.
Multi-aspect permutation methods for
cytomorphometric data under multivariate
directional alternatives with application to
comparative neuroanatomy
1
1
1
1
Livio Corain , Bruno Cozzi , Jean-Marie Graïc , Ludovica Montanucci , Luigi
2
3
1
Salmaso , Antonella Peruffo , Ruben Carvajal-Schiaffino , Enrico Grisan
1
1 University of Padova, Italy
2 University of Santiago de Chile, Chile
3 King’s College London, United Kingdom

Abstract
When bio-medical imaging shape data refer to multiple single-cell
morphological features and the goal is to inferentially compare different
populations, it appears that traditional statistical shape analysis methods are
not suitable for handling multivariate directional alternatives. This is actually a
central issue because it does not allow to draw conclusions on whether some
populations have cells that are smaller/more regular/denser vs.
larger/irregular/sparse. After organizing the neural cell descriptors in
multidimensional domains such as size, regularity and density, we propose a
data representation model in the form of a two-way multivariate linear effects
model. On the related location and scatter parameters, i.e. by using a multi-
aspect approach, and under multivariate directional alternatives, we propose
to apply the union-intersection combination-based methodology as
inferential method to separately test and rank the possible equality vs.
dominance of two or more populations. We numerically prove the
effectiveness of the proposed methodology through a simulation study where
cell shape data were obtained by simulating slices from randomly generated
geometric solids within a volume. Finally, we applied the proposed procedure
to a comparative neuroanatomy study aimed at quantifying possible
morphometric structural differences in the brain cytoarchitecture of three sex-
related bovine populations, i.e. male, female and natural intersex.

Keywords
Nonparametric combination; multivariate ranking; permutation tests

1. Introduction
Morphometrics or morphometry is a quantitative way of addressing the
shape comparisons that have always interested biologists. In neuroscience
structural differences in the brain cytoarchitecture represent the anatomical
substrate underlying the functional differences. Especially in the field of
neurodegenerative pathologies, studying structural changing in brain tissue
could be a powerful instrument to carry out morphometric analysis providing

125 | I S I W S C 2 0 1 9