Sites Inria

Version française

PARIETAL Research team

Modelling brain structure, function and variability based on high-field MRI data.

Team presentation

PARIETAL is an INRIA Research Team within the Neurospin platform of CEA Institute. It is located in Saclay, in Neurospin building.
Neurospin is a platform for the acquisition of neuroimaging data based on high fields MRI scanners, which are the most powerful ones in France today. The scanners are used to acquire high-quality data, whose resolution is also optimized (about 1mm).
Parietal aims at addressing several issues raised by the analysis of this data, in order to benefit from the potential of this data. In this perspective, Parietal is involved in the introduction of novel statistical and geometrical analysis tools which enable neuroscientists to better understand human brain structure, variability and function.
Such tools are especially needed to describe and understand the sequence of phenomena which constitute brain activity, which are now actively investigated throughout the world. They also aims at analysing inter-individual differences , with a special interest in genetic factors, which in turns has strong implications in the understanding of brain diseases.
Parietal gives a free access to its tools through the scikit learn (machine learning), joblib (scientific computing in Python), nilearn and mne-python (functional neuroimaging) software.

Research themes

  • Population imaging relates features of brain images to rich descriptions of the subjects such as behavioral and clinical assessments. We use predictive analysis pipelines to extract functional biomarkers of brain disorders from large-scale datasets of resting-state functional Magnetic Resonance Imaging (R-fMRI), Magnetoencephalography (MEG) and Electroencephalography (EEG). We also built tools for automated data analysis which facilitate processing large datasets at scale. Some of our results are highlighted below.
  • A new approach for brain analysis, called inverse inference (or brain-reading), has become recently popular. The main idea is to consider the fMRI analysis as a pattern recognition problem, i.e. using a pattern of voxels to predict a behavioral, perceptual or cognitive variable. In this approach, the accuracy of the prediction can be used to validate (or invalidate) that the pattern of voxels used in the predictive model is implied in the neural coding. In short, reverze inference is an approach for decoding neural activity.
  • Let us consider neuroimaging problems with tens of thousands to millions of brain scans (that can easily be reached when concatenating time series of brain images). Such datasets are large both in feature and sample directions. For this research axis, we contribute algorithms for data compression and online learning and have an intense activity in software development.
  • The decrease in scanning time is a crucial issue in MRI for improving patient comfort, limiting image distortions and reducing exam costs. Alternatively, MR image resolution improvement in a standard scanning time (e.g., 200 μm isotropic in 20 min) would allow neuroscientists and doc-
    tors to push the limits of their current knowledge and to significantly improve both their diagnosis and patients’ follow-up. This is the aim of compressed sensing for MRI.