Hervé Lombaert

News

Invited Talks & Seminars

Oct 2015 Spectral Random Forests CSAIL, MIT, Boston Invited Talk in the Biomedical Series Exploiting Geometry for Efficient Learning of Large Data, via Spectral Forests and Laplacian Forests Invited Talk @CSAIL, MIT
Oct 2015 Spectral Learning, Applications on Brains Martinos Center, MGH, Harvard Medical School, Boston Invited Talk in the Neuroimaging Lab Efficient Learning on Brain Surfaces, with applications on how to transfer functional maps between complex brain surfaces. Talk given during a visit at the MGH Martinos Center. Talk @Harvard Medical School
Oct 2015 Introductory Talk on Spectral Analysis MICCAI, Munich Workshop on Spectral Analysis Introductory Talk on Spectral Analysis, at MICCAI - Inaugurating the Workshop Program Organization of the Workshop
Sep 2015 Learning on Surface Data Philips Research, Paris Medisis group, Invited Talk Invited talk in the medisis group at Philips on how to learn surface data efficiently from complex datasets. Invited Talk @Philips Research
March 2015 Brain Transfer: Fast Cortical Matching University College of London, UK TIG group, Invited Talk Invited talk in the Translational Imaging Group at UCL, on how cortical data such as functional maps can be transfered from subjects to subjects. Invited Speaker @UCL
Feb 2015 Patient Numerique: Machine Learning & Imagerie Medicale Palais des Congres, Paris Microsoft Tech Talk (18,000 visitors), Invited Talk Intived talk at the Microsoft Tech Days event in Paris, in French, on Le Patient Numérique between Machine Learning and Medical Imaging. Invited Speaker @Microsoft Tech Days
Oct 2014 Laplace & Spectres Polytechnique, Montreal Computer Engineering Seminars, Invited Talk Presentation given in the Computer Engineering Dept. on Geometrical Representations of Shapes, and on Machine Learning exploiting the structure of image spaces. Invited Speaker @Polytechnique Montreal
May 2014 Prevenir plutot que guerir Microsoft France, Paris, Invited Talk Prevention et Numerique, nouvelle donne pour le public How computational models can be used for prevention. This talk is part of a three-keynote event held at Microsoft France. Keynote Speech @Microsoft France

Video here
Apr. 2014 Statistical Modeling of the Cardiac Fiber Architecture University College of London, UK Presentation made during Research meetings held at the University College of London.
March 2014 Correspondence de cortex cerebreaux, approche par graphe Ste-Justine Children's hospital, Invited Talk Prestigeous research seminars held at Ste-Justine children's Hospital in Montreal. Invited Speaker @Ste Justine Hospital, Montreal
Nov. 2013 Diffeomorphic Spectral Correspondence Imperial College of London, UK BioMedIA seminars The diffeomorphic spectral correspondence method brings smooth and invertible maps to applications demanding fast brain matching. IPMI is a select meeting in medical imaging with unlimited-timed question sessions.
July. 2013 Diffeomorphic Spectral Correspondence Asilomar, California Information Processing and Medical Imaging The diffeomorphic spectral correspondence method brings smooth and invertible maps to applications demanding fast brain matching. IPMI is a select meeting in medical imaging with unlimited-timed question sessions. Erbsmann Prize (Best Paper Award)
June. 2013 Diffeomorphic Spectral Correspondence McGill University, Montreal CREATE Program Summer School Talk on the diffeomorphic spectral correspondence method. This is the first summer school organized by the CREATE program, which is a multi university collaboration across Quebec.
Apr. 2013 Spectral Correspondence Ecole de Technologie Superieure, Montreal Spectral representations captures large deformations. This talk is based on the earlier presentation given at Microsoft Research.
Mar. 2013 Spectral Correspondence Microsoft Research, UK, Invited Talk Spectral correspondence exploits spectral graph representions of shapes in order to facilitate the challenging problem of dense point matching on meshes. The presentation focuses on how to extend these concepts from mesh matching to image registration. Invited Speaker @Microsoft Research, Cambridge, UK
Feb. 2013 Spectral Demons Auron, France Asclepios Scientific Retreat Spectral Demons enables the capture of very large image deformations by utilizing novel forces that are derived from Spectral Graph Theory. Invited Talk
Oct. 2012 Groupwise Spectral Demons Acropolis, Nice MICCAI MCV Workshop The construction of atlases from images with large deformations is challenging due to the hard recovery of such deformations. This groupwise spectral approach is one step further the efficient handling of large deformations. Best Paper Award
Jun. 2012 PhD Defense Polytechnique, Montreal Atlas Construction for Measuring the Variability of Complex Anatomical Structures. Nominated for Best Thesis
May 2012 Spectral Demons: Image registration for large deformations Centre for Intelligent Machine, McGill University, Montreal The Novel Spectral Demons algorithm allows the registration of images that have very large deformations.
Nov. 2011 FOCUSR: Fast Mesh Matching Ecole de Technologie Superieure, Montreal FOCUSR is novel method for dense vertex-to-vertex correspondence between two surface meshes. This fast, accurate, and general method is applied on a broad variaty of meshes as well as in a clinical application.
Sep. 2011 Variability of the Human Cardiac Laminar Structure Westin Harbour Conference Center, Toronto MICCAI STACOM Workshop The cardiac fibers are thought to be organized as laminar sheets. The diffusion tensor provides evidence of such secondary structure in humans, and its variability is studied along 17 myocardial segments of the left ventricle on 10 ex-vivo human hearts.
Jul. 2011 FOCUSR: Fast Mesh Matching with Spectral Correspondence CVPR group, Technische Universitat Munchen (TUM) Cortex matching is important in neuroimaging studies. We propose an approach based on spectral correspondence, where spectra of graphs derived from surface meshes are matched.
Jun. 2011 Brain Matching with Spectral Correspondence Asclepios Team, INRIA Sophia-Antipolis Cortex matching is important in neuroimaging studies. We propose an approach based on spectral correspondence, where spectra of graphs derived from surface meshes are matched.
May 2011 Statistical Analysis of the Human Cardiac Fiber Architecture from DT-MRI New York University, NYC Functional Imaging and Modeling of the Heart The statistical atlas of the cardiac fiber architecture is build for the first time in Humans. The variability of the fiber architecture is studied in 10 ex-vivo human hearts. Best Student Paper Award
May 2011 Human Statistical Atlas of Cardiac Fiber Architecture from DT-MRI Palais des Congres, Montreal Int'l Society for Magnetic Resonance in Medicine The statistical atlas of the cardiac fiber architecture is build for the first time in Humans. The construction of the atlas is automatic and the results provides the variability of the fiber architecture for 10 ex-vivo human hearts.
Aug. 2007 Landmark-based Non-rigid Registration via Graph Cuts Polytechnique, Montreal Int'l Conf. Image Analysis and Recognition Graph cuts can be used to find an optimum deformation field between two images. The use of landmarks allows user interaction, leading the optimization to favor personalized solution, and to remove ambiguities.
Nov. 2006 Registration of ICE ultrasound with DynaCT, and 4D volume compounding Polytechnique, Montreal In the first part, Samuel presented a ICE/CT registration framework. In the second part, using the ICE sequence, the reconstruction of a 4D volume is shown. with Samuel Kadoury (LIV4D, Polytechnique)
Oct. 2006 Registration using graph cut energy minimization Siemens Corporate Research The graph cut method can be used to minimize an energy. This presentation shows how this can be used with image registration.
Apr. 2006 Localisation d'instruments medicaux en IRM utilisant leur signature magnetique Polytechnique, Montreal Introducing an original idea for localization of ferromagnetic objects in MR imaging. with Arnaud Chanu (Nanorobotic Lab, Polytechnique)
Apr. 2006 Level sets: Introduction, implantation et exemples de segmentation University of Montreal Basics of the level set method, from its motion equation to its implementation with applications to segmentation.
Mar. 2006 Graph cuts: Introduction et quelques methodes de segmentation University of Montreal Basics of the graph cut algorithm and explanation of 3 related algorithms: graph cuts based active contour, lazy snapping, and multi level banded graph cuts.
Dec. 2005 Recalage par maximisation de l'information mutuelle Polytechnique, Montreal GRMSV Seminars Explanation and comparison of different voxel similarity measures such as cross-correlation, mutual information, and mutual information with gradient information. with Jeremie Theriault (LIV4D, Polytechnique)
Oct. 2005 Segmentation de structures cardiaques a l'aide de graphcuts Polytechnique, Montreal GRMSV Seminars How the graph cut algorithm can efficiently be used to segment cardiac structures.
Sep. 2005 Segmentation a l'aide de graphcuts Polytechnique, Montreal Overview of the graph cut algorithm and showing some results in cardiac segmentation.