Neuroimaging in population

Large image databases for understanding the variability of the brain and differentiate the normal from the pathological


Development, maturation and aging of the brain are influenced by many factors, both genetic and environmental, including education, diet, lifestyle, physical and intellectual activities, and incidents and illnesses occurring at during lifespan. Consequently, the morphology and the functional architecture of the brain present at the macroscopic level a very large variability between individuals. Research on the factors explaining this variability did not really begin until about thirty years ago with the advent of brain imaging techniques that can be used in healthy volunteers, such as MRI, and the implementation of a particular methodological approach, called population neuroimaging, which consists in collecting and analyzing databases containing both brain images, socio-demographic, genetic, biological and clinical data acquired from large samples of individuals.

Population neuroimaging, which aims to identify the different potential sources of brain variability and to measure the size of their effects and their interactions, is an area of research whose contributions are important from both cognitive, societal, and medical perspectives. At the cognitive level, research and our knowledge on the factors of variability of the brain remain fragmented and dependent on a particular methodological approach that has not yet totally defined its conceptual framework and its technological tools. At the societal level, population neuroimaging has an important role to play in face of the many questions and preconceptions that give rise to particularly hot debates such as for example the relationship between brain and mind, the respective roles of genes and environment on the morphology, organization, functional architecture, and performances of our brains, whether or not a sexual brain exists, the differences between left-handers and right-handers, etc. Finally, from medical and healthcare perspectives, understanding what is at the source of cerebral variability and quantifying the limits of its “normality” are of course fundamental to progress on the definition, diagnosis and development of new therapeutic approaches to cerebral diseases.


Thanks to longitudinal population-based neuroimaging studies, it is now possible to measure for each participant volume change of target brain structures. The graph above illustrates the acceleration with age of the annualized rate of atrophy of gray matter measured at the level of the hippocampus (internal cerebral structure involved especially in the mechanisms of memory) in 1,172 participants of the MRI-3-Cities study. The figure on the right shows that this acceleration is specific to the hippocampus and that the rate of atrophy of other cerebral structures does not vary with the age of the participants.


An area of research where multidisciplinary is the rule and international collaboration within large consortia a necessity

Because of the very nature of the object under study, namely the identification and quantification of potential sources of variability of the brain, and the multimodal nature of the databases to be constituted and exploited, population neuroimaging can but be multidisciplinary. Neuroimaging, epidemiology, cognitive neuroscience, neuropsychology, genetics, neurology, psychiatry and biostatistics are the main disciplines involved in a population-based neuroimaging project. For example, the IRM-3-Cités study involved neurologists, epidemiologists and imaging specialists to define the study protocol, physicians and nurses to ensure the recruitment of participants and collection of biological samples, neuropsychologists for psychometric testing, radiologists and MR technicians, computer scientists, and signal processing specialists for the acquisition and processing of MRI images, biostatisticians and geneticists for the analysis of the data and variables extracted from these MRIs.

Another key feature of the field is the high degree of international collaboration between teams. This characteristic stems from the need to assemble very large samples in order to be able to detect and quantify the small effect size of certain variability factors. Indeed, the multiple factors influencing the morphology and the functional organization of the brain have very small effects whose reproducible measurement in a brain image containing several hundreds of thousands of voxels requires a sample of several hundreds, thousands or even tens of thousands individuals. In this case, due both to budget limits and feasibility over time, it is essential that many laboratories contribute simultaneously to the creation of a meta-database, enabling meta-analyzes to be carried out. Pooling databases from different labs raises many practical issues and is done in practice thanks to a consortium, an informal but operational structure that will ensure the management of the collaborative project, in particular the standardization of the data and the analysis protocol, the circulation of results and the writing of collectively signed articles.


The GIN-IMN a pioneering team active in population neuroimaging

The GIN was among the first teams in the world to invest the field of neuroimaging in population, thanks in particular to a collaboration established in the mid-1990s with Christophe Tzourio of the INSERM team of neuroepidemiology at La Salpétrière hospital, then directed By Annick Alpérovitch, that focused on the epidemiology of cerebral aging and associated pathologies (including stroke and Alzheimer’s disease. The collaboration of the GIN with these neuroepidemiologists began with the implementation and analysis of an MRI protocol ran in 900 participants to the EVA survey conducted in Nantes and in the multicenter PROGRESS-IRM study. It was continued by the IRM-3-Cités study, a longitudinal protocol in 5,000 subjects whose data analysis is still ongoing. Thanks to the simultaneous relocation of the GIN and the unit of Christophe Tzourio in Bordeaux in 2011, this collaboration was further intensified, notably with MRi-Share, a neuroimaging protocol in a sample of 2,000 students from the University of Bordeaux, as well as with the WAIMEA project that aims to detect early white matter abnormalities and to measure their impact on brain connectivity.

The figure on the left shows that, in a cohort of 750 healthy elderly subjects from the EVA study , homozygous subjects for the ε4 allele exhibited an early hippocampus atrophy, predictive of severe cognitive decline, measured 7 years later. Conversely, no differences were observed between heterozygous and non-carrier subjects, suggesting that the impact of a single ε4 allele on the hippocampus is delayed over time. The figure on the right shows a Manhattan graph illustrating the pangenomic association of the hippocampus volume carried out within the CHARGE consortium including 9,234 elderly participants. The probability of association is given here as a function of the position of all the loci tested on the whole genome and shows that the gene variants associated with the reduction of the hippocampus volume concern genes involved in apoptosis (HRK), embryonic development (WIF1) or oxidative stress (MSR3B).


The GIN-IMN is a member of two particularly active international consortia, CHARGE (coordinated by Sudha Seshadri, Boston University) and ENIGMA (coordinated by Paul Thompson, University of Southern California), who have already done joint work to further increase power on total sample of nearly 40,000 participants. This constraint on sample size and statistical power is also the primary justification for the UK Biobank imaging study (coordinated by Paul Mattews, Imperial College, London), which is the largest population-based neuroimaging project ever undertaken on 100,000 individuals for a cost of £ 43 million.

This population-based approach to neuroimaging is also at the heart of the European MULTI-LATERAL project, coordinated by Clyde Francks (Max Planck Institute, NL) and whose GIN-IMN is a partner, that aims at identifying the anatomical, functional and genetic determinants of cerebral lateralization for language functions. Within this consortium, over 15,000 MRIs of individuals belonging to the UK-Biobank, MRI-share and BIG cohorts (COGNOMICS project, Donders Institute, NL) will be analyzed by the GIN-IMN and searched for association with their genetic profiles.

Outside the framework of this research in epidemiological neuroimaging, the GIN-IMN has also developed population-based neuroimaging approaches on its own research questions, and particularly that of hemispheric specialization. It was for this last question that the BIL&GIN base was designed and acquired on a sample of 453 individuals, of which 45% were left-handed.