Pierrick Coupé, Laboratoire Bordelais de Recherche en Informatique and Jose Vicente Manjón Herrera, Universidad Politécnica de Valencia is visiting Aarhus and will give guest talks at CFIN.

21.03.2018 | Henriette Blæsild Vuust

Pierrick Coupé, M.Sc. Ph.D.
CNRS researcher at the Laboratoire Bordelais de Recherche en Informatique
Head of PICTURA group

Title: Patch-based MRI Analysis: From voxel to knowledge

Abstract:

This talk presents the work that we have done on quantitative MR analysis, computer-aided diagnosis and brain monitoring. These topics are detailed in 3 parts. In the first part, I introduce the principle of our patch-based segmentation method and their extensions. Afterwards, the results obtained by our patch-based segmentation method are analyzed for several applications. In the second part, I show how we have extended our patch-based segmentation framework to patch-based grading of brain structures. Then, the performance of our patch-based grading method to achieve Alzheimer’s disease diagnosis and prognosis is evaluated. In the third part, I describe the tools that we developed to perform brain monitoring. First, the pipeline proposed to perform quantitative brain analysis is detailed. Second, the construction of the standard models is presented. Moreover, new medical and neuroscientific knowledge on the development and the aging of the brain produced during their estimation are discussed. Finally, the developed open access volBrain platform is described. To conclude this talk, I discuss the limitations and the perspectives of my research about patch-based MRI analysis.

Jose Vicente Manjón Herrera, M.Sc. Ph.D.
Associate Professor, Applied Physics Department
IBIME research group, Medical Imaging Area.
ITACA institute, Universidad Politécnica de Valencia.

Title: New trends in medical image processing: From patches to deep learning

Abstract:

This talk presents an overview of the current state of the art in several medical image processing problems and our current research lines. Patch-based methods represent the state of the art in medical image processing but they require hand-made tuning of their parameters which makes its development difficult. On the other hand, the deep learning technology has emerged as an efficient out of the box solution from many problems simplifying the feature extraction and parameter setting problems. In this talk, we present some preliminary examples of our work in the combination of these two technologies to solve problems such as denoising, superresolution, non-linear registration and segmentation.

ALL ARE WELCOME.