The French National Research Agency Projects for science

Voir cette page en français

ANR funded project

Emergence (Emergence)
Edition 2012


ERATRANIRMA


Fast encoding and real-time adaptation for advanced Neuro MRI applications

Fast encoding and real-time adaptation for advanced neuro-MRI applications
Fast spatial encoding of the MRI signal using non-cartesian k-space sampling and coil sensitivity parallel encoding improves advanced MRI applications. Small signal variations would be further better quantified through dynamic adaptation of acquisition paramrters to subject motion. We aim at combining fast spatial encoding and real-time adaptation approaches to open new applications in neuro-imaging.

Making application of subtle MRI techniques possible for clinical characterization of cerebral diseases
Medical imaging plays a continuously increasing role in diagnosis or therapeutic follow up of multiple pathologies. Magnetic Resonance Imaging (MRI) in particular is a multi-purpose modality combining unique features. For brain characterisation, it stands as the reference method.
With longer average lifespans, cerebral diseases related to aging increasingly become a public health problem. Cancer, dementia, multiple sclerosis, Alzheimer's and Parkinson's diseases, and acute or degenerative neuropathologies are defined by national authorities as major topics for medical research deserving strong support. Their detailed characterization heavily relies on MRI. However, recent advanced neuro-imaging techniques allowing functional or metabolic assessment of tissue often require long scan times. Cost and duration of the examination are increased, as well as the risk of data corruption from subject motion. This occurs particularly frequently with children or the elderly.
Fast spatial encoding of MR signal using non-cartesian sampling trajectories and parallel sensitivity encoding approaches can improve advanced applications like functional MRI (fMRI), arterial spin labelling (ASL) perfusion, spectroscopy and white matter fibre tracking from diffusion tensor images (DTI). Quantification of small signal variations would be further enhanced by dynamic adaptation of the acquisition parameters to head motion of the subject.
Our project has the ambition to combine fast encoding and real-time adaptation approaches, in order to accelerate transfer to the clinic of research developments and to open new neuro-imaging applications on commercially available MRI scanners.

Putting together accurate MRI scanner calibration, rigid motion navigators, high-performance GP-GPU image reconstruction techniques
The instrumental chain of magnetic field gradient waveform production for k-space spatial encoding is under particularly heavy duty in fast imaging techniques. For an optimal image reconstruction quality, it must be properly calibrated. This instrumentation aspect is one of the primary axes of the project.
Real-time adaptation of acquisition parameters to subject's head motion necessitates to very fastly determine the latter. Various methods are possible, amongst which some implying to equip each subject with a dedicated device before MRI examination. We rather target an approach avoiding such complimentary steps which lenghten examination preliminaries and lead to reduced accessible volume within magnet bore. Subject's head motion will be approximated as properly described by 6 degrees of freedom: 3 associated with translation, 3 with rotation. MRI signals will be exploited to estimate these parmeters, similarly to what is already practised in some cardiac applications.
Image reconstruction from data sampled according to non-cartesian distribution in k-space and simultaneously from coil elements featuring inhomogeneous sensitivity profiles is a subject of active research. Existing algorithms are computationnally costly, but can be sped up by using massively parallel GP-GPU computation co-processors. We develop some original aspects of such techniques and try to reach a degree of performance compatible at least with making images available to the radiologist during subject examination. This is a requirement in view of acceptance of our developments.

Results

In progress;

Expected results should come from:
- the potential to exploit the targeted characterization techniques in routine use, for cognitive science research as well as and especially for advanced clinical diagnosis of cerebral pathologies;
- the potential to apply the innovative and efficient image reconstruction algorithms developped in other fields than MRI.

Outlook

In progress.

Various cognitive science research projects will rapidly benefit from the developed techniques.

In the clinic, techniques such as BOLD fMRI, perfusion measurement by arterial spin labelling, cerebral vasoreactivity characterization, diffusion tensor measurement and white matter fibers tractography, and spectroscopy should be better performed. Examination success rate should be increased.

Scientific outputs and patents

Procédé d'estimation d'un produit de convolution / METHOD FOR ESTIMATING A CONVOLUTION PRODUCT. FR3043225 / EP3163462 / US2017146626.

Gridding: a highly parallel implementation. Lamalle L. Magn. Reson. Mater. Phys. Biol. Med. 29(1 Supplement):510, 2016. ESMRMB 2016, 33rd Annual Scientific Meeting, Vienna, AT, Sep. 29–Oct. 1, # 510.

Gridding: finite kernel support extent implies multiple local data sample weights in convolution estimation. Lamalle L. Magn. Reson. Mater. Phys. Biol. Med. 29(1 Supplement):512, 2016 / ESMRMB 2016, 33rd Annual Scientific Meeting, Vienna, AT, Sep. 29–Oct. 1, # 512

Multi-dimensional phase unwrapping: a new and efficient linear algebraic formulation using weighted least-squares. Lamalle L., Gousios G., and Urvoy M. In Proc. Intl. Soc. Mag. Reson. Med. 24, # 252, Singapore, May 7–13, 2016.

Gridding: A Highly Parallel Computing Implementation Using Locally
Determined Weights. Lamalle L. ISMRM Workshop on Data Sampling & Image Reconstruction, #38, Sedona, AZ, Jan. 17-20, 2016.

Gridding: Finite Kernel Support Extent Implies that Sampling
Density Compensation Should Be Estimated Locally. Lamalle L. ISMRM Workshop on Data Sampling & Image Reconstruction, #41, Sedona, AZ, Jan. 17-20, 2016.

Partners

CIC - CHUG Centre d'Investigation Clinique de Grenoble - Centre Hospitalier Universitaire de Grenoble

Floralis UJF Filiale

RMNBN — SFR UJF — IRMaGe Structure Fédérative de Recherche RMN biomédicale et Neurosciences — Université Joseph Fourier Grenoble 1

ANR grant: 360 246 euros
Beginning and duration: février 2013 - 24 mois

Submission abstract

Medical imaging plays a continuously increasing role in diagnosis or therapeutic follow up of multiple pathologies. Magnetic Resonance Imaging (MRI) in particular is a multi-purpose modality combining unique features. For brain characterisation, it stands as the reference method. With longer average lifespans, cerebral diseases related to aging increasingly become a public health problem. Cancer, dementia, multiple sclerosis, Alzheimer's and Parkinson's diseases, and acute or degenerative neuropathologies are defined by national authorities as major topics for medical research deserving strong support. Their detailed characterization heavily relies on MRI. However, recent advanced neuro-imaging techniques allowing functional or metabolic assessment of tissue often require long scan times. Cost and duration of the examination are increased, as well as the risk of data corruption from subject motion. This occurs particularly frequently with children or the elderly. Fast spatial encoding of MR signal using non-cartesian sampling trajectories and parallel sen- sitivity encoding approaches can improve advanced applications like functional MRI (fMRI), arterial spin labelling (ASL) perfusion, spectroscopy and white matter fibre tracking from diffusion tensor images (DTI). Quantification of small signal variations would be further enhanced by dynamic adaptation of the acquisition parameters to head motion of the subject. Our project aims at a proof of concept in the field of Technology for Health, with software Engineering aspects. It has the ambition to combine fast encoding and real-time adaptation approaches, in order to accelerate transfer to the clinic of research developments and to open new neuro-imaging applications on commercially available MRI scanners. We have built, on research equipment, expert know-how in implementing techniques for fast multi-dimensional spatial encoding associated with irregular sampling of image Fourier space. Original variants of an appropriate reconstruction algorithm and improved calibration of acquisition techniques based on this approach have been locally conceived and implemented. We propose to adapt and extend these developments to modern clinical MRI scanners, state-of-the-art notably by their parallel spatial sensitivity encoding possibilities and their potential for real time adaptation of acquisition parameters. W e will evaluate the benefit to neuro-imaging applications in research and clinical contexts. Algorithms, codes and software developed within the project will be license-protected, implying agreement prior to commercial or industrial use, and engagement to citation prior to academic exploitation. Care will be taken to protect further inventions to appear during the project. Another very important valorization target is transfer to the clinic of the advanced neuro-imaging protocols developed.

 

ANR Programme: Emergence (Emergence) 2012

Project ID: ANR-12-EMMA-0056

Project coordinator:
Monsieur Laurent LAMALLE (Structure Fédérative de Recherche RMN biomédicale et Neurosciences — Université Joseph Fourier Grenoble 1)
Laurent.Lamalle@nullujf-grenoble.fr

Project web site: http://eratranirma.univ-grenoble-alpes.fr

 

Back to the previous page

 

The project coordinator is the author of this abstract and is therefore responsible for the content of the summary. The ANR disclaims all responsibility in connection with its content.