JOB : PhD Fellowship - INRIA & Ecole Centrale de Paris - France
Vision List Digest:
Article 19,
Volume 27, Issue 4
From: "Nikos, Paragios"
Post-Followup: submission@VISLIST.com
Registration between 3D Preoperative & interventional Images
Type de poste : Doctorant
Lieu de travail : Paris-Saclay
http://www.ecp.fr
, http://www.inria.fr/saclay
Thème de recherche : Systèmes biologiques
Projet : GALEN
http://www.inria.fr/recherche/equipes/galen.en.html
RESEARCH CONTEXT:
Volume registration is one of the most critical problems in medical
image analysis. It is used to fuse information between different
modalities, as well as to assess the evolution of diseases or
therapeutic strategies. The aim is to determine a transformation that
will establish correspondences between two signals. In the most general
case, the signals are related with an unknown non-linear transformation.
Recovering this transformation involves three aspects, (i) the type of
transformation, (ii) the similarity function used to compare the two
signals, and (iii) the optimization method used to recover the lowest
potential of the designed objective function.
PHD RESEARCHER WORK DESCRIPTION:
The aim of this thesis is to address registration between pre-operative
3D annotated data and 2D interventional images coming from in-vivo
vrirtual endoscopic surgery. In such a context, the use of discrete
optimization will be considered, where one seeks for an out of plane
projection of the 2D images to the 3D volume and an in-plane non-rigid
deformation towards addressing tissue shift. Efficent techniques from
discrete optimization will be considered to address the problem. The
estimation will decomposed into a master (out of plane projection) and a
slave problem (in plane deformation). In order to account for different
modalities, a kernel-based approach will be considered first to
determine the transport/similarity function between them. Once such a
metric has been determined, the registration will be casted using a
continuous interolation model over non-uniform control points.
Required knowledge and background : B.Sc./B. Eng.
Master in Applied Mathematics and/or Machine Learning and/or medical
imaging/computer vision
Programming skills, either in MATLAB or C++, Prior experience in the
field of image registration will be a plus
Additional information & Application:
http://www.talentsplace.com/syndication1/inria/ukdoc/details.html?id=PNGFK026203F3VBQB6G68LOE1&LOV5=4509&LOV6=4516&LG=EN&Resultsperpage=20&nPostingID=2356&nPostingTargetID=5648&option=52&sort=DESC&nDepartmentID=28
Place of work: Assistance Publique Hôpitaux de Paris, Pitié Salpêtrière,
PARIS
Contact: Prof. Nikos Paragios mailto:nikos.paragios@ecp.fr
,
http://vision.mas.ecp.fr
http://www.vislist.com