Julien Pontabry

Phd candidate

ICube - MIV 300 Bd Sébastien Brant BP 10413 67412 Illkirch CEDEX - France Phone : +33 (0) 3 68 85 44 84 Office : C223 Mail : julien(dot)pontabry(at)icube(dot)fr

Research activities

Feature selection for fetal brain maturation studies

Selection of the area of huge variability on the temporal lobe (depicted by the geen color). The selected points are focused on the sulci and allow to locate the patterns of the fetal brain maturation (here, the gyration process is characterized).

Description

Le brain maturation, i.e. the way the brain structures itself during its evolution in the fetal state, is one of the major contemporary enigma. The actual knowledge is derived from post mortem studies. This difficulty is now overcomed by advanced imaging technics in vivo ; especially MRI. With some fetal images (health controlled), assumed to be a representative part of the human population, population study methods allow to highlight evolution patterns of the in utero human brain. Nevertheless, the large number of parameters is still a limitation.

A standard way to study such high-dimensional space is to reduce the dimension. There are two families of methods. The first family is the feature extraction, which aim to find a compact space suited for data representation. However, physical meaning of the new space is unclear. The second family is the feature selection. The new space is subset of the original space. Then, the physical meaning is kept and make easier to understand the maturation process.

Funding

This research are a part of a European project ERC Starting Grants on brain maturation.

With

• F. Rousseau
• M. Schweitzer
• F. Champ
• A. Bentaieb

Publications

• Sélection de caractéristiques pour l'étude de la maturation cérébrale

Software

BTK (sources, documentation)

---

Probabilistic atlas building of the fetal brain

Morphometrical analysis of the brain maturation using the jacobian projected on the grey matter surface.

Morphometrical analysis of the brain maturation using the jacobian projected on the interface between grey and whie matter.

3D model of brain tissues computed from a fetal brain population.

Coronal slices of the longitudinal model computed from a fetal brain population. Ages are plotted on the temporal axes (weeks). First row: probability map of the gray matter ; Second row: anatomical image.

Description

The human brain maturation, i.e. how the brain is structured during the fetal state, is a great modern enigma. Current knowledge is often based on post-mortem studies. However, technical advances of in-vivo acquisition systems, such as MRI, allow the studies to go beyond. Based on several clinical fetal brain images (controlled cases), supposed to be representative of the global human population, population studies are used to highlight evolution patterns of the in-utero human brain.

Atlas building is a fondamental tool for such population studies. An atlas is an average structure, which represents a population and contains the shared properties and evolution patterns of the considered population. The proposed method uses non-parametric data-driven approaches, without any prior. So, no hypothesis is made about the shape of the evolution modeling. The results are evaluation using quantitative and novel methodology to ensure the consistency of the atlas within the data.

Funding

This research are a part of a European project ERC Starting Grants on brain maturation.

With

• F. Rousseau
• M. Schweitzer

Publications

• Longitudinal Probabilistic Atlas of the Fetal Brain
• Construction d'atlas en IRM de diffusion

Software

BTK (sources, documentation)

---

Tractography with diffusion weighted MRI: application to fetal brain

Full tractography of a fetal brain. The reconstructed fibers are colored by their global orientation.

Coronal and sagital views of white matter tractography of a fetal brain (corpus callosum, pyramidal fasciculus).

Axial and coronal views of the corpus callosum connectivity map (left ang middle) and sagital view of the pyramidal fasciculus connectivity map (right) of a fetal brain.

Description

The brain is principaly composed by gray and white matters. The white matter is mainly formed by a fibrous tissue naturally caused by the axonal part of neurons. The water in the human brain is constrained by the tangles of tissues and diffuses along them, revelating their trajectories. Thus, the images formed are indirect observations of the trajectories of fiber bundles connecting one part of the brain to another. Tractography is to determine fiber paths using diffusion weighted MRI images.

For clinic usage, tractography of a diffusion weigthed MRI image of a patient aims to verifiy connections between differents parts of the human brain or to prepare a surgery. For research uses, the purposes of this technique is to learn more about brain network or to study the brain maturation (e.g. the fetal brain evolution).

The goal of tractography is to delineate trajectories along constrainted water diffusion in the brain, using noised data (the noise is due to system, environnement, etc.). The proposed solution uses trajectory tracking methods on videos (particular filter). This techniques allows an efficient solution space exploration while being robust to noise. Raw observations are modeled by spherical harmonics which allows to get a probability density function of diffusion for each point in the image.

Funding

This research are a part of a European project ERC Starting Grants on brain maturation.

With

• F. Rousseau
• E. Oubel

Publications

• Probabilistic Tractography Using Q-Ball Modeling and Particle Filtering, MICCAI 2011
• Probabilistic Tractography Using Q-Ball Modeling and Particle Filtering: Application to Adult and in-utero Fetal Brain Studies

Associated publications

• Reconstruction Methods for in-utero Fetal Brain MRI, RITS 2011
• An Open-Source Toolkit for Fetal Brain MR Image Processing, MICCAI 2011 Workshop
• BTK : An Open-Source Toolkit for Fetal Brain MR Image Processing

Software

BTK (sources, documentation)

Teaching activities (French only)

Année 2011-2012

Programmation : JavaScript (PROG4)

DUT Services et Réseaux de Communication 2ème année, IUT de Haguenau

Description

De nos jours, l'interface d'un site web se doit d'être dynamique. Le seul langage HTML ne suffit pas dans cette tâche. En plus d'être un langage orienté objet, le JavaScript permet de rendre un site web codé en HTML plus dynamique. Ce module vise à apprendre aux étudiants la manipulation d'un langage de programmation orienté objet et à appliquer ces connaissances à la conception de sites web dynamiques grâce au JavaScript.

Outils de gestion et de modélisation informatique (OMGI)

DUT Services et Réseaux de Communication 1ère année, IUT de Haguenau

Description

Les outils informatique de gestion et de modélisation de données (tableur, présentation) sont prépondérant pour la présentation efficace de données à un public. Ce module de bureautique vise à apprendre aux étudiants l'utilisation de tels outils. Le tableur OpenOffice Calc et de l'outil de présentation OpenOffice Impress sont étudiés tout au long des TP composant ce module.

Harmonisation (maths)

DUT Services et Réseaux de Communication 1ère année, IUT de Haguenau

Description

Les étudiant venant de tous horizons, il est souvent important d'harmoniser leur niveau en mathématiques. Ce module vise à leur fournir les bases essentielles de mathématiques nécessaires pour l'apprentissage de l'informatique.

Année 2010-2011

Outils de gestion et de modélisation informatique (OMGI)

DUT Services et Réseaux de Communication 1ère année, IUT de Haguenau

Description

Les outils informatique de gestion et de modélisation de données (tableur, présentation) sont prépondérant pour la présentation efficace de données à un public. Ce module de bureautique vise à apprendre aux étudiants l'utilisation de tels outils. Le tableur OpenOffice Calc et de l'outil de présentation OpenOffice Impress sont étudiés tout au long des TP composant ce module.

Programmation : JavaScript (PROG4)

DUT Services et Réseaux de Communication 2ème année, IUT de Haguenau

Description

De nos jours, l'interface d'un site web se doit d'être dynamique. Le seul langage HTML ne suffit pas dans cette tâche. En plus d'être un langage orienté objet, le JavaScript permet de rendre un site web codé en HTML plus dynamique. Ce module vise à apprendre aux étudiants la manipulation d'un langage de programmation orienté objet et à appliquer ces connaissances à la conception de sites web dynamiques grâce au JavaScript.

Publications

<anyweb> http://icube-intranet.unistra.fr/papr/appli.php?author=pontabry&title=&labo=tous&team=toutes&annee1=&annee2=&display=rap+&nationalRank=toutes&project=tous&hide=0&hide=0 </anyweb>