Probabilistic Modeling of Brain Sulci Based on Statistical
Analysis of 3D Surfaces
C. Barillot1, P. Hellier1, G. Le Goualher2
and B. Gibaud2
1: IRISA, INRIA/CNRS Unit, VISTA project, Rennes, France, 2: UPRES-EA 2232 " Cortex Cérébral et Epilepsies
", Rennes, France
Goal and Rationale
One of the major problems in functional neuroimaging is to match anatomical
and functional data from different
subjects. Usually this task is performed on morphological basis coming
from MRI. This work intends to tackle this problem by computing statistical
models of cortical sulci from analytical representations of these sulci
obtained automatically from 3D MRI using the " active ribbon " method [1].
Our goal is to use these " local " statistical anatomical models as a substrate
to compare functional recordings coming from different subjects (e.g. MEG
or fMRI). This statistical modeling of cortical sulci will allows the description
of their shapes and their variability and can be used as constraints to
assist non-linear registration of human brains by inter-individual matching
of anatomy following similar ideas than those proposed in [2, 3]. We propose
to apply to neuro-anatomy a general statistical framework defined
for modeling déformable object [4, 5]. The model proposed here is
devoted to be used for digital brain atlases.
[1] G. Le Goualher, C. Barillot, Y. Bizais,
Int. J. of Pattern Recognition and Artificial Intelligence, 1997,
11:8
[2] Collins D.L., Le Goualher G., Venugopal
R., Caramanos A., Evans A.C., Barillot C, Lecture Notes in Computer
Sciences: Visualization in Biomedical Computing,
1996, Vol.1131:307-316.
[3] Thompson P.M., Toga A.W., Medical Image
Analysis, 4(1):271-294.
[4] Cootes T., Cooper D., Taylor C., Graham
J., Image and Vision Computing, 1992, Vol.10(5):289-294.
[5] Kervrann C., Heitz F, Proc. of IEEE
Computer Vision & Pattern Recognition, 1994, pp.724-728
