Advancement of Field-Monitored Reconstructions of Non-Cartesian Diffusion MRI at Ultra-High Field
Abstract
Diffusion magnetic resonance imaging (dMRI) is a non-invasive imaging modality that is sensitive to tissue microstructure. Its ability to characterize microstructural abnormalities makes it an effective diagnostic tool, particularly for the detection of neurological disorders. However, dMRI suffers from inherently low signal, making it challenging to achieve robust measurements at high spatial resolutions. This issue is especially prominent in human imaging and when employing advanced dMRI techniques of higher sensitivity. Additionally, dMRI is highly susceptible to image artefacts, which is further exacerbated when implementing strategies designed to overcome low signal outputs, such as ultra-high field imaging, stronger diffusion gradients, and spiral acquisition trajectories. Field monitoring, a technique that measures field perturbations during acquisitions using an array of field probes, offers a solution by enabling the correction of artefacts and the production of high-quality diffusion data. However, given the relative recency of field monitoring development, its effective application in demanding scanning scenarios has not been extensively explored. Accordingly, the objective of this thesis work was to investigate the use of concurrent field monitoring for dMRI applications on a head-only 7T scanner and in highly undersampled acquisitions, and provide solutions for effective field-monitored reconstructions in these challenging scenarios.
It was previously observed that higher-order field monitoring is hindered when field probes extend beyond the imaging region and encounter rapid spatial field variations, leading to erroneous images. The first project aimed to address this issue by modifying the fitting algorithm to prioritize lower order fitting and less distant probes, resulting in improved characterization of field perturbations. This was further improved by introducing a second strategy that compresses field perturbation terms into fewer components, allowing the characterization of more rapid spatial variations using the standard number of field probes. Lastly, this work demonstrated the denoising benefits when incorporating compressed sensing in field-monitored reconstructions compared with conventional least squares reconstruction methods, enabling the acquisition of images in lower signal regimes.
In summary, this thesis work introduces alternative field monitoring reconstruction strategies with the goal of enhancing generalizability for expanded field monitoring usage, thereby improving dMRI sensitivity and accuracy in applications such as neuroimaging studies.