<> "The repository administrator has not yet configured an RDF license."^^ .
<> .
.
"Affine Registration of label maps in Label Space"^^ .
"Two key aspects of coupled multi-object shape\r\nanalysis and atlas generation are the choice of representation\r\nand subsequent registration methods used to align the sample\r\nset. For example, a typical brain image can be labeled into\r\nthree structures: grey matter, white matter and cerebrospinal\r\nfluid. Many manipulations such as interpolation, transformation,\r\nsmoothing, or registration need to be performed on these images\r\nbefore they can be used in further analysis. Current techniques\r\nfor such analysis tend to trade off performance between the two\r\ntasks, performing well for one task but developing problems when\r\nused for the other.\r\nThis article proposes to use a representation that is both\r\nflexible and well suited for both tasks. We propose to map object\r\nlabels to vertices of a regular simplex, e.g. the unit interval for\r\ntwo labels, a triangle for three labels, a tetrahedron for four\r\nlabels, etc. This representation, which is routinely used in fuzzy\r\nclassification, is ideally suited for representing and registering\r\nmultiple shapes. On closer examination, this representation\r\nreveals several desirable properties: algebraic operations may\r\nbe done directly, label uncertainty is expressed as a weighted\r\nmixture of labels (probabilistic interpretation), interpolation is\r\nunbiased toward any label or the background, and registration\r\nmay be performed directly.\r\nWe demonstrate these properties by using label space in a gradient\r\ndescent based registration scheme to obtain a probabilistic\r\natlas. While straightforward, this iterative method is very slow,\r\ncould get stuck in local minima, and depends heavily on the initial\r\nconditions. To address these issues, two fast methods are proposed\r\nwhich serve as coarse registration schemes following which the\r\niterative descent method can be used to refine the results. Further,\r\nwe derive an analytical formulation for direct computation of the\r\n\"group mean\" from the parameters of pairwise registration of all\r\nthe images in the sample set. We show results on richly labeled\r\n2D and 3D data sets."^^ .
"2010-06-06" .
.
"2" .
"4" .
.
"Journal of Computing, https://sites.google.com/site/journalofcomputing/"^^ .
.
.
"Journal of Computing, Volume 2, Issue 4, April 2010, https://sites.google.com/site/journalofcomputing/"^^ .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
"James"^^ .
"Malcolm"^^ .
"James Malcolm"^^ .
.
"Martha E"^^ .
"Shenton"^^ .
"Martha E Shenton"^^ .
.
"Sylvain"^^ .
"Bouix"^^ .
"Sylvain Bouix"^^ .
.
"Yogesh"^^ .
"Rathi"^^ .
"Yogesh Rathi"^^ .
.
"Allen"^^ .
"Tannenbaum"^^ .
"Allen Tannenbaum"^^ .
.
.
.
.
.
"Affine Registration of label maps in Label Space (PDF)"^^ .
.
.
.
.
.
.
.
.
.
.
.
"RegistrationLabelSpace.pdf"^^ .
.
.
"Affine Registration of label maps in Label Space (Image (JPEG))"^^ .
.
.
.
.
.
"preview.jpg"^^ .
.
.
"Affine Registration of label maps in Label Space (Indexer Terms)"^^ .
.
.
.
.