Université de Genève

Neuroscience center

Imprimer cette page
image

Group leader: Dimitri Van De Ville

Group name: Medical Image Processing Laboratory (MIPlab)

Affiliation: University of Geneva & EPFL

Research activities:

image Our main research objective is to develop new methodologies for medical image processing. Most of our projects are related to neuroimaging, including functional magnetic resonance imaging (fMRI), laser Doppler imaging (LDI), and electroencephalography (EEG). Sophisticated tools in signal processing and statistics are required to fully exploit the potential of functional brain imaging data, which are usually large, complex, and noisy. Among those tools, the wavelet transform receives our particular attention. Despite significant breakthroughs in this field, general-purpose wavelets have been used up to now, and significant gain can be expected by tailoring wavelets to the properties of the signals under investigation. For example, we study "activelets" that exploit the hemodynamic nature of the signal (such as the BOLD response in fMRI) and provide us with a more sparse representation (which is useful to derive stronger statistical measures). Currently, fMRI data analysis is performed voxel-wise and does not allow to exploit spatial coupling. We propose to develop multivariate analyses based on machine learning techniques that can take advantage of subtle coupling between voxels and lead to backward inference; so-called "mind reading" based on fMRI data. Another research axis pursues better integration of analysis methods for intrinsic and evoked brain activity. Existing methodologies are very different according to the type of activity they want to process. Statistical methods such as blind source separation have been utilized to describe intrinsic brain activity while general linear models are the workhorse to tackle evoked activity. Our point-of-view is to consider intrinsic activity as an essential element that modulates evoked activity, for example through (low-frequency) fluctuations in brain networks. We investigate into a new methodology that leads to an integrated approach. One of our primary research goals is to bridge the gap between theoretical advances and applications in neurosciences and medical imaging. To that aim, we collaborate closely with other laboratories in the Faculty of Medicine and the Geneva Neuroscience Center.

Group website: http://miplab.unige.ch/

Selected Publications:

  • T. Ethofer, D. Van De Ville, K. Scherer & P. Vuilleumier. Decoding of Emotional Information in Voice-Sensitive Cortices Current Biology vol. 19 :(12) pp. 1028-1033, 2009.
  • A. Raabe, D. Van De Ville, et al. Laser Doppler Imaging for Intraoperative Human Brain Mapping, NeuroImage vol. 44 :(4) pp. 1284-1289, 2009.
  • D. Van De Ville, M. Unser. False Discovery Rate for Wavelet-Based Statistical Parametric Mapping, IEEE Journal of Selected Topics in Signal Processing, vol. 2 :(6) pp. 897-906, 2008.
  • D. Van De Ville, M. Seghier, F. Lazeyras, T. Blu & M. Unser. WSPM: Wavelet-Based Statistical Parametric Mapping, NeuroImage vol. 37 :(4) pp. 1205-1217, 2007.

Contact:
Department of Radiology and Medical Informatics
University of Geneva
Email: dimitri (dot) vandeville (at) unige (dot) ch