MRI Signal analysis.
Magnetic resonance is a widespread medical imaging technique. Compared to other techniques, like CT or PET, it does not use ionizing radiation to obtain images; data are acquired using radiofrequency waves, static magnetic fields ad magnetic field gradients. For this reason there are not known health risks for patients.
High resolution images can be acquired and many contrast mechanisms can be employed. Combining this factor, MRI can be used either for anatomical or functional informations.
Classical whole-body scanners have a close bore where the patient lies, with serious problems for patient comfort. In recent years, some manufacturer have introduced open, C-shaped field systems. The main limitation of open magnets is a low magnetic field strength, due to the technical difficulty to obtain a strong, homogeneous field on such geometry.
The U-shaped magnet design is a novel approach to open system. Its main advantage is the possibility for the patient to stand or sit inside the scanner. This acquisition mode allows to investigate pathologies that are hidden when the patient lies down, to study articulations while bearing weights, and to stimulate fMRI volunteers with a wider range of stimuli. Sitting and standing position bring a drawback with them. Patient can not avoid small movements of the body, causing motion artefacts, blurring and other problems on image quality.
In this context our group is developing acquisition sequences and reconstruction algorithms to optimise image quality and scan time at low field. The target is to exploit all the opportunities connected with U-shaped magnet. In this context the group is currently active on these topics:
a) Phased array reconstruction on open systems.
b) Echo Planar Imaging at low field.
c) Fast imaging using non-Cartesian sampling and Compressed Sensing.
Phased-array reconstruction on open systems
A surface coil has an high signal-to-noise ratio (SNR), but it decays fast as the signal source is moved far from the coil. The result is a small field of view (FOV). Many surface coils can be combined in a phased array. The SNR of this kind of array is the same of a surface coil, and the FOV is the superposition of those of its elements.
Low field systems can benefit most from the high SNR of phased-array coils. The main limit of this kind of receivers lays in the intrinsic inhomogeneity of surface coils. In whole-body, closed scanners the inhomogeneity, or sensitivity map, is measured using a quadrature body coil. Using this map the images are corrected an recombined to obtain a full-FOV, homogeneous image. In open-field MRI system such method is not feasible.
We developed an algorithm to estimate the sensitivity maps, without using a quadrature coil. The maps are fitted from the phased-array data. When images from the elements of the array are combined our algorithm optimises the combination to avoid noise amplification in regions where none of the coils has good SNR.
Echo Planar Imaging at low-field, open MRI system
Echo Planar Imaging (EPI) sequence is one of the fastest acquisition method in MRI. With modern gradients, EPI is capable to form a complete 2D image in a few tens of milliseconds. On the other hand EPI is prone to a variety of artifacts. It is sensitive to susceptibility effects and field nonuniformities that can lead to ghosts and geometric distortion. Our studies investigate the feasibility of EPI imaging on low field open MR system, (MrOpen). This open MRI system with U-shaped magnet and field intensity of 0.5 T is able to scan patients in seated and standing position. The goal will be to perform an functional MRI (fMRI) experiment in sitting position. Such physiological position allows to have a new prospective for fMRI applications. A different view to investigate functional activation with new tasks will be developed.
Fast imaging using non-Cartesian sampling and Compressed Sensing
Non-Cartesian sampling trajectories, like radial, spiral and random sampling, provide great flexibility and some advantages compared to conventional sampling. Motion and aliasing artefact show lesser coherence, thus allowing motion-insensitive acquisitions and fast imaging even without parallel imaging. Since each trajectory always samples high and low frequencies, they are also more efficient for dynamic imaging.
Non-Cartesian sampling can be combined with Compressed Sensing (CS) to achieve very short acquisition times. The principle behind CS is the same of image and video compression. The information of an images is much lower than its actual size, therefore only a number of samples proportional to the information content is required for optimal reconstruction. For CS purposes, data must be sampled with a higher density where the signal energy is higher (where most of the information is), and this is exactly how a non-Cartesian trajectory does.
Our goal is to perform fast, motion-insensitive scans on upright patients using a low field open system. We also plan to perform dynamic imaging of moving joints using these techniques.