Septiembre 2018: Charla organizada por la Unidad Académica en Ciencia, Tecnología y Salud de la UC3M

Título: 4D reconstruction of spectral CT data with a Variable Metric proximal algorithm
Ponente: Yannik Boursier (Aix-Marseille Université)
Fecha: jueves, 27 de septiembre de 2018
Hora:   16:00-18:00
Aula:    4.0.D01 Edif. Torres Quevedo (Leganés)

Esta charla está organizada por la Unidad Académica en Ciencia, Tecnología y Salud de la UC3M, creada con el objetivo de atraer a los mejores investigadores en ciencias, ingeniería, y biomedicina para constituir un entorno cooperativo de investigación multidisciplinar. 

A dicha Unidad Académica pertenecen nuestros compañeros el Dr. Manuel Desco, el Dr. Juan José Vaquero, el Dr. Javier Pascau, la Dra. Mónica Abella y la Dra. Arrate Muñoz.

A través de un enfoque centrado en la colaboración científica entre grupos de excelencia de ámbitos complementarios y en compartir recursos, se crea un ambiente de trabajo innovador para la generación de conocimiento, la formación de investigadores y el emprendimiento en el entorno socio-sanitario.

Resumen:

Spectral CT have raised an incredible interest in the last decade due to the advent of photon counting detectors. The imXgam team at CPPM (Centre de Physique des Particules de Marseille) has developed a spectral micro-CT called PIXSCAN-FLI equipped with the XPAD3 camera also developed at CPPM for in-vivo imaging purpose. The proof-of-concept first of K-edge imaging and then of component separation has been done until in-vivo imaging.

We recently developed new algorithms in order to simultaneously separate and reconstruct the physical components of an object observed in the context of spectral Computed Tomography. Our contribution considers the underlying polychromatic model of the X-ray beam and combines it with a prior on the components of the object to reconstruct. The simultaneous separation and reconstruction of components is done by solving a non-convex ill-posed inverse problem with a variable metric proximal algorithm. The technological developments at CPPM for spectral CT will be presented and I will then focus on the methodology and the proposed algorithm. Finally, I will show the promising results recently obtained on 3D real data acquired on the PIXSCAN-FLI with regularizations that encompass in particular the positivity of the quantities of interest.

Información sobre el ponente:

Yannick Boursier graduated from the Institut Supérieur de l'Aéronautique et de l'Espace(ISAE - ENSICA/SUPAERO) and obtained the M.S degree of Applied Mathematics from the Université Paul Sabatier and ISAE, Toulouse, France, in 2004. He defended in 2007 the PhD degree from the Université Aix-Marseille 3, France under the joint supervision of P. Lamy (Laboratoire d’Astrophysique de Marseille) and F. Goudail (Institut d’Optique Graduate School, Laboratoire Charles Fabry).

He then spent a year as a research engineer at OAMP/LAM lab. (2008) and has been a postdoctoral researcher at EPFL at LTS2 and LTS4,Lausanne, Switzerland, with Prof. P. Vandergheynst and Prof. P. Frossard (2009).

Since 2009, he works at the Centre de Physique des Particules de Marseille (CPPM) as an Assistant/Associate Professor (Maître de Conférences) at Aix-Marseille Université, France.

His research field concerns the mathematical aspects of signal, image and data processing. He is mostly interested in sparse representation of signals and their use in solving inverse problems in image/signal processing, including compressed sensing and computer vision.

He is interested in developing convex and non-convex optimization tools and applying them to real data. Currently, his favourite application subjects are multimodal and spectral-tomographic imaging in the context of the development of the hybrid pixels based X-ray detectors, a technological breakthrough.

He published 12 journal papers and 16 publications as international conference proceedings and gave more than 35 talks in international conferences or workshops.