Alice Dufour
Phd candidate
|
 |
Research activities
PhD : Segmentation and modeling of cerebrovascular structures in 3D medical imaging.
Advisor: Christian Ronse
Co-adivsor: Joseph Baruthio
Supervisor:Nicolas Passat
Vascular segmentation in 3D angiographic images
The segmentation of vascular structures consist in extract the relevant information from the angiographic data, to allow radiologist to focus only on the information analysis. Works upstream and based on mathematical morphology are fitting on the angiographic data segmentation. They are been improve and applied on our data.
- Segmentation of the brain arterial network
Low resolution image (In gree : true positive, in yellow : false negative et in red : false positive). 
High resolution image (In gree : true positive, in yellow : false negative et in red : false positive). 
3D visualisation of the segmentation of Willis circle.
Collaborations : Benoît Naegel, Olena Tankyevych, Hugues Talbot and Valérie Wolff
Cerebro-vascular atlas
To integrate anatomical knowledge in the segmentation process, the knowledge should be modelling. The second goal of this thesis consist to develop vascular models, and the generating pipeline.
- Cerebro-vascular atlas
Atlas of vessel probability density
Vein (blue)/artery(red) probability.
Vessel size. From blue 0.5mm to red 5mm.
Collaborations : Olena Tankyevych and Hugues Talbot
Master : Filtering vascular flow artifacts in fMRI.
Supervisor: Daniel Gounot
Co-Supervisor:Nicolas Passat
Report
Problematic
Since 1990s the field of magnetic resonance imaging (MRI) have evolved to enable the study of brain functional activity, this is named the fonctional MRI (fMRI). The localization of the activated areas is determined by statistical calculation and is function of the considered paradigm1. That localization is based on the BOLD effect (Blood Oxygen Level Dependent) which induces the local increase of MRI signal. Nevertheless, these variations are very low (around 1%). It is the reason why it is essential to reduce the noise contained by fMRI images such as patient’s movements, respiratory or cardiac noise. Moreover, [Dagli, NeuroImage, 1999] have demonstrated that the heartbeats in the vascular network induce change into the fMRI signal. The aim of these works is determining the influence of the brain vascularization in fMRI, in order to reduce it. The study of brain vascularity is based on anatomical knowledge obtained by brain angiography atlases.
Contribution
These works introduce a first approach of the vascular atlas, used to improve the detection of brain activity. The first step of these works is creating a cerebral vascular atlas with the required information, that is to say only the intracranial vessels network. The second step is to determine the influence of vascular NMR signal in the activity detection. This part of work is difficult and takes a lot of time, because of the number of way to determine this influence. Last but not least, this report present the creation of a toolbox, which permit to reduce the noise caused by cardiac signals in cerebral vascular network.
Teaching activities
2012-2013: IUT Robert Schuman
- P32: Abstract data type, and structure (Java).
- S22: Undestand and know the systems(Shell, python).
2011-2012: UFR Math-Info
- L2 Conputer Science: Computer architecture. Assembly langage.
- L2 Math-Physics-Chemical: Applied programming. Introduction to Octave.
2010-2011: UFR Math-Info
- L2 Math-Physics-Chemical: Applied programming. Introduction to Octave.
- L3 Math: Oriented Object Programming
2009-2010: UFR Math-Info
- L2 Math-Physics-Chemical: Applied programming. Introduction to Octave.
- L3 Biology: Algorithm and data structure. Introduction to C.