LREN Principal Investigator
Maitre d'Enseignement et de Recherche
phone: +41 21 314 9593
Laboratoire de Recherche en Neuroimagerie - LREN
Departement des neurosciences cliniques
Centre Hospitalier Universitaire Vaudois (CHUV)
Mont Paisible 16
I am a Senior Lecturer at the University of Lausanne and vice director of the Laboratoire de Recherche en Neuroimagerie (LREN) of Departement des Neurosciences Cliniques (DNC) at the University Hospital of Lausanne (CHUV). I obtained my PhD in neuroscience at the university Pierre et Marie Curie, Paris. I was research fellow at MRC-CBSU in Cambridge and then at the Welcome center for neuroimaging in London before my arrival in Lausanne in 2010. I am coordinating and directing the work of the various components of the medical informatics platform of the Human Brain Project. The aims are to create a facility for medical ICT, federate very large volumes of data from European hospitals about diseases of the brain (imaging, data from tissue and blood samples, clinical data, medical histories, genomic data, etc.) and finally use this data to discover biological signatures of diseases. I have several years of experience in multidisciplinary research with deep knowledge and expertise in two key domains mathematical/statistical modelling and in neuroscience/psychology. He used functional imaging to probe cognitive function and used my mathematical background to test new hypotheses pertaining the explanation of individual differences.
The goal of my teaching activities is to educate students and collaborators to the advances neuroimaging tools. The teaching include training to basic statics, advanced multivariate approach, Bayesian models and machine learning. Beyond classical teaching my motivation is to engaged to the student using active learning methods based on demonstration with real examples to individual or small groups. In my teaching I provide theoretical knowledge and practical tutorial of the statistical packages MATLAB or SPSS and to dedicated neuroimaging software like Statistical Parametric Mapping (SPM). I teach mainly at the doctoral level at the Lemanic neuroscience doctoral school. I also organize and teach twice a year a SPM course that attract students in Lausanne and outside Lausanne. During this period, this course became a reference for neuroimaging courses with an international reputation. I organize twice a month a method clinics and method for dummies sessions open to all students and researchers involved in neuroimaging.
Fundamental Research in Clinical Neurosciences Through Neuroimaging
My main interest in neuroimaging lies in the study of inter-individual differences: the understanding of the nature and extent of inter-subject variation is critical for understanding the neural basis of cognitive processes in normal and abnormal populations. My overarching goal is to provide new tools for a more accurate and comprehensive investigation of neural processes in both normal and clinical populations.
Comprehensive Outcome Models of Aphasia. The goal of the study is to build a predictive model of patients with aphasia after brain injury. The specific aims of the study are: To monitor longitudinal anatomical and neuropsychological changes in sample of patients with ischemic stroke and aphasia. 2) To use language functional mapping in well-recovered patients to test alternative recovery hypotheses. 3) To use multivariate methods based on anatomo-functional features in order to build a predictive model of patients outcomes. During the past year we have setup the necessary logistic in order to ensure patients’ data collection and to guarantee successful completion of this project. We invite participants (submission approved by the local ethics committee) from all consecutive patients who were admitted to the stroke unit or intensive care unit of the Centre Hospitalier Universitaire Vaudois (CHUV) with a main diagnosis of acute ischemic stroke. More than 10 patients per week (> 500 patients/year) are routinely admitted.
Brain networks of cognition and personality in Alzheimer’s disease. The aim of this project is to explain variability in clinical and anatomical patterns and the effects of individual differences in pre-clinical stages of AD in order to develop a more inclusive anatomo-functional model of AD. Our specific aims are: 1) To test the significance of individual differences factors in explaining the observed variability in anatomical changes in Mild Cognitive Impaired (MCI) patients as well as in healthy controls. 2) To include brain regions related to these individual differences and to provide a more accurate classification of patients in the pre-clinical stages with a better prediction of disease’s outcome. 3) To derive functional markers of memory components and to provide a functional mapping of the critical regions affected by AD, in particular the hippocampus, entorhinal cortex and the perirhinal cortex, regions that are otherwise difficult to delineate on the basis of anatomical landmarks.
Model of Memory. Recent interactive models of memory systems highlight that, to understand the functions of the sub-regions of the MTL it is necessary to identify the sources of interaction with the other memory nodes in the brain. In this study we used fMRI and multiple cues probabilistic learning (MCPL) task to probe the different memory system. We adapted the Weather prediction task2 to a virtual game environment and used pseudo-letters to render the cue integration task similar to a pseudo-word learning paradigm. We further expand the methods used in previous studies by 1) using network analyses to delineate the key memory nodes and 2) using computational model of cue weights utilization3 to take into account the individual differences in learning strategy. We expect that the division of labor for performing the task will result in activation of the hippocampus (episodic memory), fusiform gyrus (perceptual memory), pre-frontal cortex (working memory) and basal ganglia (procedural memory). Our hypothesis is that interactions (cooperation and/or competition) between the memory nodes explain the temporal dynamic of the activations and predict trial-by-trial subject's performance.
Big data and biological signature of Alzheimer’s disease. The current study is to identify homogeneous groups of patients, characterised by a set of parameterised latent causes, which constitute what we choose to call "disease signatures". The advantage of the disease signatures is that they are based on mechanistic, deterministic and predictive rules as opposed to purely descriptive (phenomenological or clinical) features. A successful parsing of a large population by disease signature will have major implications for the future of disease classification and provide preliminary data to support the Human Brain Project proposal to the European Community. The objectives are 1) Characterise the target population of Alzheimer’s patients using brain structure information from Magnetic Resonance Imaging (MRI) as well as genotyping, neuropsychological, biological and clinical measures. 2) Scale up a pilot study (200 subjects) to a largescale level (c.a. 2-4000 subjects) with credible partners (Sanofi-Aventis and the Bordeaux Team - Orgogozo, Dartigues, Tzourio & Mazoyer). 3) Apply novel data mining approaches on High Performance Computing facilities (EPFL and others) to extract disease signatures. 4) Demonstrate that the rule-based disease-signatures provide comprehensive models that explain the variability among patients and aged normal subjects in a reduced multidimensional space. By post hoc deep phenotyping, biochemical and molecular genetic analysis and appropriate pharmacological target identification impact drug discovery methods.
Applied Reseach in Statistical Sciences, Bioinformatics and Medical Informatics
Methodological platform. The purpose of this project is to provide methodological tools for the research community not only for the university but also beyond. We have already constructed processing pipeline for big data and state-of the art toolbox. In particular, we created first tools for automated MRI multivariate and machine learning analyses.
Neuroimaging computing Platform. The purpose of this project we are building a state of the art neuroimaging computing platform next to the MRI scanner and data acquisition platform of LREN. The goal of the computing platform is to enable the use of neuroimaging data acquired at LREN but also the smooth combination with the genetic data acquired by the BIL (Biobanque institutionnelle). We working on this project with CHUV ITRC, the UNIL IT and the EPFL. We are now one of the main user of the High-Computing Cluster created at CHUV.
Medical informatics Platform. CHUV is leading this part of the human brain project. The work focus on building a platform for federating data that have been collected in enormous quantities around the world. With sufficient provenance surrounding these data we will enable analytic strategies with unprecedented sensitivity. Furthermore, we will identify acquisition parameters that provide the best sensitivity and reliability. Coupled with sophisticated informatics and statistical tools, we will characterize the human brain across subjects and lifetimes.
Kherif F, Josse G, Price CJ. Automatic top-down processing explains common
left occipito-temporal responses to visual words and objects. Cereb Cortex. 2011
Ramponi C, Barnard PJ, Kherif F, Henson RN. Voluntary explicit versus
involuntary conceptual memory are associated with dissociable fMRI responses in
hippocampus, amygdala, and parietal cortex for emotional and neutral word pairs.
J Cogn Neurosci. 2011 Aug.
Draganski B, Ashburner J, Hutton C, Kherif F, Frackowiak RS, Helms G, Weiskopf
N. Regional specificity of MRI contrast parameter changes in normal ageing
revealed by voxel-based quantification (VBQ). Neuroimage. 2011 Apr
Davis MH, Ford MA, Kherif F, Johnsrude IS. Does Semantic Context Benefit
Speech Understanding through "Top-Down" Processes? Evidence from Time-resolved
Sparse fMRI. J Cogn Neurosci. 2011 Jul 11.
Seghier ML, Kherif F, Josse G, Price CJ. Regional and hemispheric determinants
of language laterality: implications for preoperative fMRI. Hum Brain Mapp. 2011
Ali N, Green DW, Kherif F, Devlin JT, Price CJ. The role of the left head of
caudate in suppressing irrelevant words. J Cogn Neurosci. 2010