Research Assistant Professor in the OHSU Center for Spatial System Biomedicine (OCSSB), Biomedical Engineering department (BME).
My work is applied to all medical and biomedical imaging modalities:
- Electron Microscopy (EM). Automatic segmentation of the different elements inside and outside the cell as the nuclei membrane, nucleolus, the collagen, etc. using deep learning.
- Immuno-Fluorescence (IF). Automatic segmentation of nuclei coupled with cell characterization and quantifications.
- H&E. Automatic segmentation and classification of nuclei, using clustering and deep learning.
- MRI. Early prediction of advanced breast cancer response to chemotherapy.
Post doctorate (2013-2014)
Post doctoral position in the Center for Spoken Language Understanding (CSLU) and OHSU Center for Spatial System Biomedicine (OCSSB).
Supervisor: Izhak Shafran.
I worked biomedical image like Electron Microscopy (EM) and Immuno-Fluorescence (IF), mainly on immunogold detection, cell segmentation and characterization.
Post doctorate (2011-2012)
Post doctoral position in the Center for Mathematical Morphology (CMM) into the « Ecole des Mines de Paris » (ENSMP / Mines-ParisTech) for TeleOphta project: pattern recognition, computer aided diagnosis, image processing and management.
Supervisor: Etienne DECENCIERE.
Leading Xiwei Zhang’s PhD.
Partners: AP-HP, LaTIM, ADCIS company.
The goal of TeleOphta is the automatic detection of diabetic retinopathy in eye fundus images (or more precisely, the detection of retinopathy absence).
Context: the AP-HP has created a telemedicine network in order to facilitate the follow-up of diabetic patients, because the number of diabetics increases (an estimation states that 30% of american people will become diabetic in 2030 and 50% in 2050) but the number of ophthalmologists decreases (it is sometimes necessary to wait 12 months for an appointment). So patients go in a screening center, two images of each eye are taken, images are sent via a server to ophthalmologists and an opinion is returned. But this network generates thousands of images each month and 65% of these images are healthy, so do not require any analysis. So we want to automate the detection of healthy images in order to get time saving.
The first step is the detection of all main structures such as vessels, papilla and fovea, and specific pathologies of diabetic retinopathy: exudates, micro-aneurisms, cotton wools and neo-vessels.
Post doctorate (2009-2010)
Post doctoral position in the Center for Mathematical Morphology (CMM) into the « Ecole des Mines de Paris » (ENSMP / Mines-ParisTech) for RAMIS project: pattern recognition, computer aided diagnosis and image processing.
Supervisor: Jesus ANGULO.
Context: Cell phenotyping.
Partners: Pierre Fabre laboratories, CBIO, ADCIS company.
Context: pharmaceutical laboratories are trying to discover new molecules in order to elaborate treatments against cancer. The molecules must kill cancerous cells and usually they block cancerous cell in a specific phase of mitosis. So RAMIS goal, was to automatically determine in which mitosis phase the cells are.
This work has first required to preprocess stacks of images acquired from confocal microscope: z-projection (new algorithm particularly efficient), to correct illumination and segmentation. Next we extract characteristics of shape and texture (to elaborate new ones too) and then we classify cells with learning algorithms.
Three more contributions:
- Fly wings characterization and classification for the evaluation of the impact of genetic mutations, for IBDM lab.
- Characterization and classification of dermis images (3D stacks acquired from confocal microscope) in order to estimate the age of patients.
The final goal of L’Oréal Company was to evaluate the impact of treatments in dermatology.
- Authentication of signatures in order to avoid falsifications of bank checks, for ICDAR 2010 challenge in collaboration with the Dutch scientific police.
PhD Thesis (2005-2009)
Title: Shape and texture indexes, application to cell nuclei classification.
Supervisors: Jean-Luc MARI, Bernard FERTIL, Jean Sequeira. Defended on June 18 2009.
Design new shape indexes and a new statistical method to characterize textures.
Master SIS training course (2005)
Elaboration of new tools for virtual sculpt.
Supervisor: Romain RAFFIN.
Here are my report and a publication.
TER (2004)
Voici mon Travail d'Etude et de Recherche (Ter) de maîtrise : « détection d'objets ajoutés dans une scène ».
Il fût encadré par J-M BOI de l'équipe I&M (ex LXAO).