|
|
|
| e-Pub |
Previous | 
Home
Section: New Results
Research axis 1: Medical Image Computing in Neuroimaging
Extraction and exploitation of complex imaging biomarkers involve an imaging processing workflow that can be quite complex. This goes from image physics and image acquisition, image processing for quality control and enhancement, image analysis for features extraction and image fusion up to the final application which intends to demonstrate the capability of the image processing workflow to issue sensitive and specific markers of a given pathology. In this context, our objectives in the recent period were directed toward 4 major methodological topics:
Diffusion imaging
Optimal selection of diffusion-weighting gradient waveforms using compressed sensing and dictionary learning
Participants : Raphaël Truffet, Emmanuel Caruyer.
Acquisition sequences in diffusion MRI rely on the use of time-dependent magnetic field gradients. Each gradient waveform encodes a diffusion weighted measure; a large number of such measurements are necessary for the in vivo reconstruction of microstructure parameters. We proposed here a method to select only a subset of the measurements while being able to predict the unseen data using compressed sensing. We learnt a dictionary using a training dataset generated with Monte-Carlo simulations; we then compare two different heuristics to select the measures to use for the prediction. We found that an undersampling strategy limiting the redundancy of the measures allows for a more accurate reconstruction when compared with random undersampling with similar sampling rate [57].
A Bayes Hilbert Space for Compartment Model Computing in Diffusion MRI
Participant : O. Commowick.
The single diffusion tensor model for mapping the brain white matter microstructure has long been criticized as providing sensitive yet non-specific clinical biomarkers for neurodegenerative diseases because (i) voxels in diffusion images actually contain more than one homogeneous tissue population and (ii) diffusion in a single homogeneous tissue can be non-Gaussian. Analytic models for compartmental diffusion signals have thus naturally emerged but there is surprisingly little for processing such images (estimation, smoothing, registration, atlas-ing, statistical analysis). We propose to embed these signals into a Bayes Hilbert space that we properly define and motivate. This provides a unified framework for compartment diffusion image computing. Experiments show that (i) interpolation in Bayes space features improved robustness to noise compared to the widely used log-Euclidean space for tensors and (ii) it is possible to trace complex key pathways such as the pyramidal tract using basic deterministic tractography thanks to the combined use of Bayes interpolation and multi-compartment diffusion models [26]
This work was done in collaboration with A. Stamm, A. Menafoglio and S.K. Warfield.
Arterial Spin Labeling
Patch-Based Super-Resolution of Arterial Spin Labeling Magnetic Resonance Images
Participants : Cédric Meurée, Pierre Maurel, Jean-Christophe Ferré, Christian Barillot.
Arterial spin labeling is a magnetic resonance perfusion imaging technique that, while providing results comparable to methods currently considered as more standard concerning the quantification of the cerebral blood flow, is subject to limitations related to its low signal-to-noise ratio and low resolution. In this work, we investigated the relevance of using a non-local patch-based super-resolution method driven by a high-resolution structural image to increase the level of details in arterial spin labeling images. This method was evaluated by comparison with other image dimension increasing techniques on a simulated dataset, on images of healthy subjects and on images of subjects diagnosed with brain tumors, who had a dynamic susceptibility contrast acquisition. The influence of an increase of ASL images resolution on partial volume effects was also investigated in this work. [56]
The development of this super-resolution algorithm in the context of a thesis financed by Siemens Healthineers conducted to a stay of one month of the PhD candidate in Erlangen, during summer 2018. This immersion into the neuro-development team allowed to integrate the proposed solution with tools in use within this team. Part of the work also consisted in reducing the calculation time, a factor of 5 being achieved at the end of these four weeks.
Resting-state ASL : Toward an optimal sequence duration
Participants : Corentin Vallée, Pierre Maurel, Isabelle Corouge, Christian Barillot.
Resting-state functional Arterial Spin Labeling (rs-fASL) in clinical daily practice and academic research stay discreet compared to resting-state BOLD. However, by giving direct access to cerebral blood flow maps, rs-fASL leads to significant clinical subject scaled application as CBF can be considered as a biomarker in common neuropathology. Our work here focused on the link between overall quality of rs-fASL and duration of acquisition. To this end, we consider subject self-Default Mode Network (DMN), and assess DMN quality depletion compared to a gold standard DMN depending on the duration of acquisition [46].
Quantitative imaging
Identification of Gadolinium contrast enhanced regions in MS lesions using brain tissue microstructure information obtained from diffusion and T2 relaxometry MRI
Participants : S. Chatterjee, O. Commowick, C. Barillot.
A multiple sclerosis (MS) lesion at an early stage undergoes active blood brain barrier (BBB) breakdown. Identifying MS lesions in a patient which are undergoing active BBB breakdown is of critical importance for MS burden evaluation and treatment planning. However in non-contrast enhanced structural magnetic resonance imaging (MRI) the regions of the lesion undergoing active BBB breakdown cannot be distinguished from the other parts of the lesion. Hence gadolinium (Gd) contrast enhanced T1-weighted MR images are used for this task. However some side effects of Gd injection into patients have been increasingly reported recently. The BBB breakdown is reflected by the condition of tissue microstructure such as increased inflammation, presence of higher extra-cellular matter and debris. We have thus proposed a framework to predict enhancing regions in MS lesions using tissue microstructure information derived from T2 relaxometry and diffusion MRI (dMRI) multicompartment models. We showed that combination of the dMRI and T2 relaxometry microstructure information can distinguish the Gd enhancing lesion regions from the other regions in MS lesions [23].
A three year follow-up study of gadolinium enhanced and non-enhanced regions in multiple sclerosis lesions using a multi-compartment relaxometry model
Participants : S. Chatterjee, O. Commowick, B. Combes, C. Barillot.
Demyelination, axonal damage and inflammation are critical indicators of the onset and progress of neurodegenerative diseases such as multiple sclerosis (MS) in patients. Due to physical limitations of imaging such as acquisition time and imaging resolution, a voxel in a MR image is heterogeneous in terms of tissue microstructure such as myelin, axons, intra and extra cellular fluids and free water. We present a multi-compartment tissue model which estimates the water fraction (WF) of tissues with short, medium and high relaxation times in a relaxometry MRI voxel. The proposed method is validated on test-retest data of healthy controls. This model was then used to study longitudinal trends of the tissue microstructures for two sub-regions of the lesions: gadolinium enhanced (E+) and non-enhanced (L−) regions of MS lesions in 10 MS patients over a period of three years. The water fraction values in E+ and L− regions were found to be significantly different (p < 0.05) over the period of first three months. The results of this study also showed that the estimates of the proposed relaxometry model on brain tissue microstructures have potential to distinguish between regions undergoing active blood brain barrier breakdown from the other regions of the lesion [49].
This work was done in collaboration with Onur Afacan and Simon K. Warfield from Harvard Medical School.
Multi-Compartment Model of Brain Tissues from T2 Relaxometry MRI Using Gamma Distribution
Participants : S. Chatterjee, O. Commowick, C. Barillot.
The brain microstructure, especially myelinated axons and free fluids, may provide useful insight into brain neurodegenerative diseases such as multiple sclerosis (MS). These may be distinguished based on their transverse relaxation times which can be measured using relaxometry MRI. However, due to physical limitations on achievable resolution, each voxel contains a combination of these tissues, rendering the estimation complex. We presented a novel multi-compartment (MCT2) estimation based on variable projection, applicable to any MCT2 microstructure model. We derived this estimation for a three-gamma distribution model. We validated our framework on synthetic data and illustrated its potential on healthy volunteer and MS patient data [32].
This work was done in collaboration with Onur Afacn and Simon K. Warfield.
A 3-year follow-up study of enhancing and non-enhancing multiple sclerosis (MS) lesions in MS patients demonstrating clinically isolated syndrome (CIS) using a multi-compartment relaxometry (MCT2) model
Participants : S. Chatterjee, O. Commowick, B. Combes, A. Kerbrat, C. Barillot.
Obtaining information on condition of tissue microstructures (such as myelin, intra/extra cellular cells, free water) can provide important insights into MS lesion. However, MRI voxels are heterogeneous in terms of tissue microstructure due to the limited imaging resolution owing to existing physical limitations of MRI scanners. Here we evaluated a multi-compartment relaxometry model and then used it to study the evolution of enhancing (USPIO and gadolinium positive) and non-enhancing lesions in 6 MS patients with CIS characteristics over a period 3 years with 7 follow-up scans post baseline [31].
This work was done in collaboration with Onur Afacn and Simon K. Warfield.
Atlases
Anisotropic similarity, a constrained affine transformation: Application to brain development analysis
Participants : A. Legouhy, O. Commowick, C. Barillot.
The study of brain development provides insights in the normal trend of brain evolution and enables early detection of abnormalities. We proposed a method to quantify brain growth in three arbitrary orthogonal directions of the brain through linear registration. We introduced a 9 degrees of freedom transformation that gives the opportunity to extract scaling factors describing brain growth along those directions by registering a database of subjects in a common basis. We applied this framework to create longitudinal curves of scaling ratios along fixed orthogonal directions from 0 to 16 years highlighting anisotropic brain development [39].
This work was done in collaboration with François Rousseau under the ANR MAIA project.
Simultaneous EEG/fMRI
Automated Electrodes Detection during simultaneous EEG/fMRI
Participants : M. Fleury, C. Barillot, E. Bannier, P. Maurel.
The coupling of Electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) enables the measure of brain activity at high spatial and temporal resolution. The localisation of EEG sources depends on several parameters including the knowledge of the position of the electrodes on the scalp. An accurate knowledge about this information is important for source reconstruction. Currently, when acquiring EEG and fMRI together, the position of the electrodes is generally estimated according to fiducial points by using a template. In the context of simultaneous EEG/fMRI acquisition, a natural idea is to use magnetic resonance (MR) images to localise EEG electrodes. However, most MR compatible electrodes are built to be almost invisible on MR Images. Taking advantage of a recently proposed Ultra short Echo Time (UTE) sequence, we introduce a fully automatic method to detect and label those electrodes in MR images. Our method was tested on 8 subjects wearing a 64-channel EEG cap. This automated method showed an average detection accuracy of 94% and the average position error was 3.1 mm. These results suggest that the proposed method has potential for determining the position of the electrodes during simultaneous EEG/fMRI acquisition with a very light cost procedure" [11], [35].
This work was done in collaboration with Marsel Mano from Biotrial.
Inference in neuroimaging
Validity of summary statistics-based mixed-effects group fMRI
Participant : Camille Maumet.
Statistical analysis of multi-subject functional Magnetic Resonance Imaging (fMRI) data is traditionally done using either: 1) a mixed-effects GLM (MFX GLM) where within-subject variance estimates are used and incorporated into per-subject weights or 2) a random-effects General linear model (GLM) (RFX GLM) where within-subject variance estimates are not used. Both approaches are implemented and available in major neuroimaging software packages including: SPM (MFX analysis; 2nd-Level statistics), FSL (FLAME; OLS) and AFNI (3dMEMA; 3dttest++). While MFX GLM provides the most efficient statistical estimate, its properties are only guaranteed in large samples, and it has been shown that RFX GLM is a valid alternative for one-sample group analyses in fMRI [1]. We recently showed that MFX GLM for image-based meta-analysis could lead to invalid results in small-samples. We investigated whether this issue also affects group fMRI [42].
This work was done in collaboration with Prof. Thomas Nichols from the Oxford Big Data Institute, UK.
Choosing a practical and valid Image-Based Meta-Analysis
Participant : Camille Maumet.
Meta-analysis provides a quantitative approach to summarise the rich functional Magnetic Resonance Imaging (fMRI) literature. When image data is available for each study, the optimal approach is to perform an Image-Based Meta-Analysis (IBMA) [1]. A number of IBMA approaches have been proposed including combination of standardised statistics (Z's), just effect estimates (E's) or both effect estimates and their standard errors (SE's). While using both E’s & SE’s and estimating between-study variance should be optimal, the methods are not guaranteed to work for small number of studies. Also, often only standardised estimates are shared, reducing the possible meta-analytic approaches. Finally, because the BOLD signal is non-quantitative care has to be taken in order to insure that E's are expressed in the same units [2,3], especially when combining data from different software packages. Given the growing interest in data sharing in the neuroimaging community there is a need to identify what is the minimal data to be shared in order to allow for future IBMAs. We investigated the validity of 8 IBMA approaches [41].
This work was done in collaboration with Prof. Thomas Nichols from the Oxford Big Data Institute, UK.
Machine learning
Learning sparse predictor from hybrid EEG-fMRI neurofeedback
Participants : C. Cury, C. Barillot, P. Maurel.
Electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) both allow measurement of brain activity for neuro-feedback (NF), respectively with high temporal resolution for EEG and high spatial resolution for fMRI. Using simultaneously fMRI and EEG for NF training is very promising to devise brain rehabilitation protocols, however performing NF-fMRI is costly, exhausting and time consuming, and cannot be repeated too many times for the same subject. The original contribution of this work concerns the prediction of NF scores from EEG recordings only, using a training phase where both EEG and fMRI NF are available. We have proposed a model able to predict NF scores from coupling EEG-fMRI (NF-EEG-fMRI) of 17 subjects in motor imagery task. The prediction of NF-EEG-fMRI scores was found as satisfactory with a significant improved performance with respect to what EEG can provide alone, when adding to NF-EEG, the prediction of NF-fMRI from EEG signals. The prediction of NF-fMRI significantly adds information and increases the quality of the estimated NF-EEG-fMRI scores.
This work was done in collaboration with Remi Gribonval from the Inria/IRISA PANAMA team.