Section: New Results
Group inference and comparison of neuroimaging data with genetic data
Participants : Bertrand Thirion, Jean-Baptiste Poline, Merlin Keller, Cécilia Damon.
A Bayesian model for brain activity detection
To overcome the limitations of standard voxel-based testing methods, such as Statistical Parametric Mapping (SPM), we have introduced a new approach for fMRI group data analysis. This is a region-based procedure that aims at outlining global networks instead of local (voxel-based) analysis. Using a Bayesian model selection framework, the functional network associated with a certain cognitive task is selected according to the posterior probabilities of mean region activations, given a pre-defined anatomical parcellation of the brain. This approach enables us to control a Bayesian risk that balances false positives and false negatives, unlike the SPM-like approach, which only controls false positives. On data from a mental calculation experiment, it detected the functional network known to be involved in number processing, whereas the SPM-like approach either swelled or missed the different activation regions.
First joint analysis of neuroimaging and genetic data
Imaging genetic studies linking functional MRI data and Single Nucleotide Polyphormisms (SNPs) data may face a dire multiple comparisons issue. In the genome dimension, genotyping DNA chips allow to record of several hundred thousands values per subject, while in the imaging dimension an fMRI volume may contain 50k voxels. Finding the brain and genome regions that may be involved in this link entails a huge number of hypotheses, hence a drastic correction of the statistical significance of pairwise relationships, which in turn reduces crucially the sensitivity of statistical procedures that aims at detecting the association. It is therefore desirable to set up as sensitive techniques as possible to explore where in the brain and where in the genome a significant link can be detected, while correcting for family-wise multiple comparisons (controlling for false positive rate).
In neuroimaging, the problem has been addressed during the past last 15 years with a numerous methods and software. The most popular tests developed in neuroimaging are testing in a statistical map for either the voxel intensity, or the size or mass of clusters defined by thresholding the statistical map, with permutation-based statistical validation. In the analysis of SNP data, a number of techniques have been designed as well. Most are based on the idea that the combination of p-values found at adjacent SNPs will be more significant and more biologically relevant than considering the SNPs independently.
We are working, in collaboration with V.Frouin (CEA, Neurospin) on a simple test for imaging genetic data (voxel x SNPs) based on the idea that we wish to detect contiguous brain regions linked to neighbour SNPs on the genome. The method detects clusters defined by a threshold in the product (four-dimensional) data-set, and calibrates the null hypothesis using permutations. While computationally intensive, the technique is conceptually simple, corrects for the multiple comparisons in both the imaging and the genetic dimensions, accounts for the spatial structure of the data (correlation of the imaging data and the linkage disequilibrium of the genetic data. We are currently evaluating this approach on a data-set that includes 94 subjects for which fMRI and SNP data.