Registration remains a central issue for detecting and quantifying shape variations. In
this talk, I survey some of the past and current registration methods developed in our
research group Epidaure at INRIA, and show how they can be combined with some
specific methods to detect and quantify shape variations, either from intra-patient
longitudinal studies, or from inter-subject medical images. Applications cover the followup
of brain diseases, multimodal image fusion for image-guided neuro-surgery,
construction of 3-D brain representations from 2-D optical cross-sections or
autoradiographies, segmentation of deep grey nuclei from MR images, modeling of
sulcal variability through tensor maps. I indicate some promising directions, including the
processing of in vivo cellular images.
All references are available on the web site of Epidaure Laboratory:
http://www-sop.inria.fr/epidaure/BIBLIO/
Audio (MP3 File, Podcast Ready)