MRI sequence development for robust and time efficient blood velocity quantification simultaneously in the heart and great vessels
Recrutement en cours/passé: 
Recrutement en cours

The Goal is to design, implement and evaluate an MR imaging sequence dedicated to the simultaneous quantification of blood flow velocities in the heart and the great vessels. The acquisition should be robust in terms of geometry planning and sensitivity to physiological motion, and efficient and clinically feasible in terms of acquisition time.

Context: MR imaging can measure blood flows in vivo in four dimensions (4D = space + time) and therefore identify abnormal blood flow related with cardiovascular diseases. It is now clear that abnormal blood flow has an important role in the evolution of these diseases. However, the current limitations of the technique, such as the long scan time, have prevented its use in a clinical setting.

The 4D measurement of the intracardiac velocity field requires 15 to 20 minutes because of the need to repeat the measurement at least 4 times and to use prospective respiratory gating, which reduces scan efficiency significantly.

Challenges to be addressed: Advances in spatio-temporal acceleration techniques and non-Cartesian k-space sampling have enabled important reductions in scan times. In this direction, we explored a combination of 3D radial sampling and compressed-sensing algorithms for under-sampled data reconstruction. It demonstrated several advantages, such as improved spatial resolution and improved volume coverage without penalty on scan time. In addition, self-navigation and respiratory motion correction, allowed for 100% scan efficiency (1–3). Furthermore, the large volume coverage of this approach is clinically relevant in the context of cardiovascular diseases. However, the limited dynamic range of the velocity measurement reduces the impact of this advantage.

MR-based velocity measurements require the pre-selection of a maximum expected velocity (or velocity sensitivity, VENC). Adjusting the VENC for high velocities, such as those expected in the aorta (~150 cm/s), reduces the velocity to noise ratio (VNR) and thus the measurement accuracy for low velocities, such as those in the pulmonary veins and the atria (~ 60 cm/s). Adjusting the VENC to low velocities improves the VNR but results in aliasing artifacts for high velocities that may be difficult to correct (Figure1).

This limitation can be addressed by acquiring several VENC values (multi-VENC), followed by a post-processing step to extract a single velocity dataset with high VNR for a wider dynamic range. The acquisition of multiple VENC values leads to a further increase in scan duration, thus making its acquisition impractical in a clinical setting. Proposed dual-VENC strategies have been based on the use of a low and a high VENC that result in significantly different eddy current effect, which velocity measurements are sensitive to. Particularly for radial sampling, inherently sensitive to eddy-current related errors, this approach is not adapted. In addition, quantification errors remain an important issue for non-cartesian k-space sampling.

Work program: The aim of the project is to address the limitations of 4D velocity measurement in MRI presented above by designing, implementing and evaluating a sequence specifically dedicated to the simultaneous quantification of blood flow in the entire heart and in the great vessels based on radial sampling.

There are three main aspects to the project: simulation, implementation/acquisition and reconstruction. The candidate will first have to identify the most relevant approaches in the literature, implement them in an MR simulator (such as JEMRIS) and evaluate their performances. This part of the project will be done in collaboration with the Montpellier group that recently developed a CFD-based 4D flow MRI simulator (4). We will extend this approach to non-Cartesian k-space sampling. This approach is novel and the importance is two-fold: it will help better understand the eddy-current related effects for 3D radial and it should help propose novel schemes for multiple velocity encodings for time efficiency and reduction in acquisition related errors.

The most suited approach will then be implemented on a Philips MRI system (Radiology Department of the Hospices Civils de Lyon, Hôpital de la Croix Rousse). This phase will consist of validating the velocity measurements on flow phantoms, using pulsatile flow conditions, and on healthy volunteers for which comparison will be made with standard sequences. Finally, image reconstruction algorithms from under-sampled measurements will need to be investigated. Both iterative and deep learning based approaches can be explored. The candidate will rely on current algorithms explored by the team and will explore strategies to include hemodynamic constraints in the reconstruction algorithm.

Upon validation, the new developed sequence will be used to explore the blood flow patterns in cardiovascular pathologies that will be identified in interaction with the Radiology Department of the HCL.


Application: Candidates with a strong background in signal/image processing methodologies with possible previous experience in MR imaging will be most suited. But we will evaluate all candidates having a masters’ degree with a profile fitting the subject. Candidates should send a letter of motivation, CV, and the grades obtained during the last two years to:

Monica Sigovan, CR CNRS,

Damien Garcia, CR INSERM,


1.            Boccalini S, Mousseaux M, Chevalier P, Boussel L, Douek P, Sigovan M. Investigation of Left-Atrial flows using a 3D radial based self-gated respiratory motion corrected 4D Flow MRI sequence. ESMRM B. 2021.

2.            Sigovan M, Schneider T, Cruz G, et al. Respiratory-resolved self-gated 3D radial 4D flow MRI: Initial results. SMRA. Stellenbosch, South Africa; 2017.

3.            Sigovan M, Duchateau N, Douek P, Prieto C, Boussel L. Velocity-based cardiac self-gating in free-running radial 4D Flow MRI. SMRA. 2021.

4.            Puiseux T, Sewonu A, Moreno R, Mendez S, Nicoud F. Numerical simulation of time-resolved 3D phase-contrast magnetic resonance imaging. Borazjani I, editor. PLOS ONE. 2021;16(3):e0248816. doi: 10.1371/journal.pone.0248816.