Grégoire Malandain

	INRIA senior research scientist
	Morpheme team (I3S/INRIA/iBV)
	Adress: Laboratoire I3S BP 121, 2000 Route des Lucioles, 06903 Sophia-Antipolis Cedex, FRANCE Email: gregoire.malandain@inria.fr Tel: +33 4 89 15 43 21

Short bio

Grégoire Malandain received an engineer degree from the Ecole Centrale des Arts et Manufactures in 1989, a PhD degree in medical image analysis in 1992, and a habilitation degree in 2006. He entered INRIA at Sophia-Antipolis in 1993, and became senior research scientist in 2001. From 1998, he was team coordinator of Epidaure team and then from Asclepios team from 2005. From 2009 to 2011, he was deputy scientific director of INRIA in charge of the "Computational Sciences for Biology, Medicine and the Environment domain".

Grégoire Malandain began research in medical image analysis in 1989 at INRIA Rocquencourt. His methodological research interests include discrete topology, image segmentation, image registration, and atlas building. Driving applications include multiple sclerosis biomarkers, radiotherapy planification, and pre-clinical and biological image analysis. His scientific interests currently focuses on biological image analysis, with applications in morphogenesis and embryogenesis.

Some selected topics/publications

Embryogenesis and morphogenesis

• Contact area-dependent cell communication and the morphological invariance of ascidian embryogenesis, Léo Guignard, Ulla-Maj Fiuza, Bruno Leggio, Julien Laussu, Emmanuel Faure, Gaël Michelin, Kilian Biasuz, Lars Hufnagel, Grégoire Malandain, Christophe Godin, Patrick Lemaire, Science, 2020, 369 (6500), pp.158. DOI, article (pdf), supplementary material [121 pages] (pdf)

  Segmentation of temporal series of 3D light-sheet microscopy images, including fusion of multiple images at each timepoint, segmentation propagation from one timepoint to the next with simultaneouslineage extraction. This article points out the importance of contact surfaces in cell communication in ascidian embryo growth (Phallusia mammillata).

• Imaging plant growth in 4D: robust tissue reconstruction and lineaging at cell resolution, R. Fernandez, Pradeep Das, V. Mirabet, E. Moscardi, Jan Traas, J.-L. Verdeil, Grégoire Malandain, Christophe Godin Nature Methods, 2010, 7, pp.547-553. DOI, article (pdf)

  Our princeps article on the segmentation of temporal series of 3D confocal microscopy images, including fusion of multiple images at each timepoint, independent segmentation of each timepoint, and lineage extraction. The main application is the quantitative analysis of the development dynamics of the floral meristem (Arabidopsis thaliana). Results have been used to build a meristem atlas

• Related PhDs
  • Manuel Petit (defended in june 2023), lineage estimation in temporal series of 3D microscopy images, inter-individual image comparison, atlas construction.
  • Gaël Michelin (defended in october 2016), Cell membrane enhancement in 3D microscopic images, co-registration of 3D+t series of microscopy images, atlas construction
  • Léo Guignard (defended in december 2015), segmentation by propagation of temporal series of 3D microscopy images, cell-to-cell communication modelling
  • Romain Fernandez (defended in november 2010), segmentation of temporal series of 3D microscopy images and lineage estimation

Atlas-based segmentation for radiotherapy planning

• Atlas-Based Delineation of Lymph Node Levels in Head and Neck Computed Tomography Images, Olivier Commowick, Vincent Grégoire, Grégoire Malandain Radiotherapy & Oncology, 2008, 87 (2), pp.281-289. DOI, article (pdf)

  One (of the many) articles on organ at risk delineation for radiotherapy planning (organs must received a controlled amount of radiation, to avoid undesirable side effects). Professor V. Grégoire is one of the most renowned specialist of head and neck tumor radiotherapist. Apart the methodological results, the article demonstrates the difficulty to deal with clinical databases that exhibits quite a lot of imperfections. Here, we had to accomodate with missing informations: all structures of interest are not delineated in all the database images.

• Related PhDs
  • Liliane Ramus (defended in july 2011), Atlases for automated delineation of organs-at-rick in a radiation therapy context.
  • Olivier Commowick (defended in february 2007), Atlases for automated delineation of organs-at-rick in a radiation therapy context.
  • Pierre-Yves Bondiau (MD) (defended in november 2004), Evaluation of image registration techniques in a radiation therapy context. Atlas-based organ-at-risk delineation has been introduced during this PhD. Pierre-Yves Bondiau is now head of the radiation therapy department of one (out of 18) french center devoted to fight cancer.

Motion compensated tomographic reconstruction

• Reconstruction of Coronary Arteries from a Single Rotational X-Ray Projection Sequence, Christophe Blondel, Grégoire Malandain, Régis Vaillant, Nicholas Ayache IEEE Transactions on Medical Imaging, 2006, 25 (5), pp.653--663. DOI, article (pdf)

  Tomography consists in reconstructing a volume (or a cross-section) from its projections. An underlying assumption is that the imaged objects are still during the acquisition of the projections, so that the projections are the ones of the same scene from different points of views.

  In this work, we perform the reconstruction of a 3D object (the coronary arteries) from X-ray projections of the beating heart acquired with a moving C-arm, ie with a deforming object. It is quite a methodogical tour de force, that incorporates methods issued from both the computer vision and the medical imaging fields.

• Model-based respiratory motion compensation for emission tomography image reconstruction, Mauricio Antonio Reyes Aguirre, Grégoire Malandain, Pierre Malick Koulibaly, Miguel Ángel González Ballester, Jacques Darcourt, Physics in Medicine and Biology, 2007, 52 (12), pp.3579-600 DOI, article (pdf)

  In this work, we addressed the motion-compensation tomographic reconstruction in a nuclear medicine context for lung tumor detection and assessment. There are several differences with the previous work: tomographic reconstruction has to be done with probabilistic methods; local deformations have to be compensated for; it is unfortunately not possible to estimate the deformation from the projections, since projections are integrated over the motion.

• Related PhDs
  • Mauricio A. Reyes (defended in december 2005) Motion-compensation tomographic reconstruction in nuclear medicine.
  • Christophe Blondel (defended in march 2004) Motion-compensated tomographic reconstruction of coronary arteries from one single rotating C-arm acquisition

Codes

Maintained codes

• Adaptive Segmentation and Tracking of Embryonic Cells (Astec). A first version has been developed by Léo Guignard during his PhD. This suite allows to process temporal series of light-sheet microscopy images, including
  • fusion of multiple acquisition
  • Segmentation and correction of one time point
  • Segmentation propagation from one time point and lineage computation
  • Drift compensation
  • Segmentation post-correction
  This suite provides the results analyzed in Contact area-dependent cell communication and the morphological invariance of ascidian embryogenesis.
  • Documentation: astec.gitlabpages.inria.fr/astec/
  • Code: gitlab.inria.fr/astec/astec
• Ascidian: a collection of tools to exploit astec results in the context of ascidian embryogenesis. This suite allows to
  • name cells of one time point of a temporal series (from a database of named embryos)
  • propagate cell names in a temporal series from one named time point (using a database of named embryos)
  • compute some statistics from a database of embryos
  This suite provides the results analyzed in Contact area-dependent cell communication and the morphological invariance of ascidian embryogenesis.
  • Documentation: astec.gitlabpages.inria.fr/ascidian/
  • Code: gitlab.inria.fr/astec/ascidian

Oldies

• 2D/3D edge detector: a small library (written in C) allowing 2D or 3D edge detection with recursive filtering is available for non-commercial uses only.
• Connected components extraction: with the edge detection library comes already a hysteresis thresholding software. Other calls (counting and labeling connected components in 2D or in 3D) are available. A small library, allowing connected components labeling and hysteresis thresholding is also available for non-commercial uses only. Designing other procedures (extracting components containing seeds, or holes filling) should be really easy.
• Resampling of 2D/3D images: Code for resampling of 2D or 3D images with linear (or nearest neighbor) interpolation using a 4 by 4 matrix is also available for non-commercial uses only.
• Computation of optimal Chamfer masks and coefficients

PhD students

Current PhD students

• Haydar Jammoul. From oct. 2023. Development variability in ascidian embryos

Former PhD students (defense year)

• Manuel Petit (2023), enc. 50%. Calcul de lignage dans des séries temporelles d'images microscopiques 3D, outil de comparaison inter-individus. Thèse en co-encadrement avec l'EP Mosaic (Lyon).
• Emmanuelle Poulain (2019), enc. 50%. Recalage déformable 2D/3D de données vasculaires avec des contraintes topologiques. Thèse Cifre avec la société General Electric Healthcare.
• Gaël Michelin (2016), enc. 100%. Rehaussement de membranes, outils de comparaison de séries temporelles d'images microscopiques 3D.
• Léo Guignard (2015), enc. 30 %. Outil de segmentation de séries temporelles d'images microscopiques d'embryon en développement. Thèse en co-encadrement avec l'EP Virtual Plants (Montpellier) et le CRBM (Montpellier).
• Thomas Benseghir (2015), enc. 50%. Recalage rigide 2D/3D de données vasculaires avec des contraintes topologiques. Thèse Cifre avec la société General Electric Healthcare.
• Marine Breuilly (2013), enc. 100%. Reconstruction tomographique de données 3D+t chez le petit animal avec une caméra couplée TEMP/CT, et évaluation de la distribution d'iode.
• Liliane Ramus (2011), enc. 100%. Construction d'atlas pour le contourage automatique d'organes à risques dans un contexte de planification de radiothérapie. Thèse Cifre avec la société Dosisoft.
• Romain Fernandez (2010), enc. 40%. Segmentation et le suivi d'images de microscopie confocale de méristèmes (parties en croissance des plantes). Thèse en co-encadrement avec l'EPI Virtual Plants et le CIRAD.
• Jean-Christophe Souplet (2009), enc. 80%. Segmentation automatique de lésions de sclérose en plaques (SEP) en imagerie par résonance magnétique (IRM) multimodale.
• Arnaud Charnoz (2007), enc. 40%. Segmentation et la mise en correspondance d'arbres vasculaires hépatiques.
• Olivier Commowick (2007), enc. 100%. Construction d'atlas pour le contourage automatique d'organes à risques dans un contexte de planification de radiothérapie. Thèse Cifre avec la société Dosisoft.
• Guillaume Dugas-Phocion (2006), enc. 80%. Segmentation automatique de lésions de sclérose en plaques (SEP) en imagerie par résonance magnétique (IRM) multimodale.
• Céline Fouard (2005), enc. 100%. Reconstruction et la segmentation de mosaïques d'images de microscopie confocale. Thèse Cifre avec la société TGS.
• Mauricio A. Reyes (2005), enc. 80%. Reconstruction tomographique d'image de médecine nucléaire avec compensation du mouvement respiratoire.
• Christophe Blondel (2004), enc. 50%. Reconstruction tomographique d'image des artères coronaires à partir d'images radiologiques avec compensation du mouvement respiratoire. Thèse Cifre avec la société General Electric Healthcare.
• Pierre-Yves Bondiau (2004), enc. 100%. Utilisation des outils de recalage d'images dans un contexte de radiothérapie. Pierre-Yves Bondiau est radio-thérapeuthe au centre Antoine Lacassagne (centre de lutte contre le cancer), et faisait son doctorat en parallèle de son activité clinique.
• Jonathan Stoeckel (2003), enc. 60%. Outils de classification pour l'aide au diagnostic en médecine nucléaire.
• Sébastien Ourselin (2002), enc. 40%. Recalage robuste d'image par la mise en correspondance de régions.
• David Rey (2002), enc. 40%. Détection de processus évolutifs dans des images médicales, avec application à la sclérose en plaques.
• Octave Migneco (2001), enc. 100%. Outils de recalage et de traitement statistique pour des images de médecine nucléaire. Octave Migneco était MCU-PH à la faculté de médecine / au CHU Pasteur de Nice, et faisait son doctorat en parallèle de son activité clinique.
• Alexis Roche (2000), enc. 50%. Mesures de similarité pour le recalage des images médicales.
• Karl Krissian (1999), enc. 50%. Segmentation d'arbres vasculaires dans des images volumiques.
• Luc Soler (1998), enc. 40%. Segmentation des tumeurs et des segments anatomiques du foie.
• Sara Fernández-Vidal (1996), enc. 80%. Amincissement topologique et de segmentation d'arbres vasculaires.