Magnetic resonance imaging (MRI) is a pivotal tool for the non-invasive examination of anatomical structures and physiological processes in the brain and body. MRI relaxometry is a
branch of MRI that enables a detailed investigation of brain tissue microstructures. Specifically, estimates of the relaxation rate of the transverse magnetization R2* (=1/T2*) and magnetic
susceptibility (‘quantitative susceptibility mapping’- QSM) are biomarkers of iron and myelin content within brain tissue. R2* and QSM estimates are highly sensitive to cardiac-induced
noise, which reduces their sensitivity to tissue content, obscuring subtle tissue changes ob
served in neurological disorders. In MRI relaxometry, the multi-echo acquisition of a single image can take several minutes, spanning multiple heartbeats. This extended duration can
lead to aliasing cardiac-induced noise, making its effect extremely challenging to address. The main objective of this PhD work was to investigate novel approaches for the acquisition
of multi-echo data, mitigating cardiac-induced noise and decreasing the variability of R2* maps and QSM.
The first part of this project involved an extensive characterization of cardiac-induced noise on 3D multi-echo MRI k-space data and R2* maps, paving the way for the creation of new sampling strategies to mitigate such noise. The second part of the project
involves the creation of three sampling strategies based on the properties of cardiac-induced noise. A sampling strategy corresponds to the order in which the raw space data are acquired. By altering such order, the impact of cardiac-induced noise on the R2* and QSM estimates can be strongly mitigated. The first strategy acquires multiple k-space data where cardiac-induced noise is the highest using a Cartesian pseudo-spiral sampling. Compared to standard linear sampling, pseudo-spiral sampling reduces the variability of R2* and QSM by respectively 22-28% and 16-19%, at a cost of a scan time increase of 14%. The second strategy instead synchronized the
data acquisition with the participant’s heart rate in real-time but did not lead to significant improvements. The third strategy, named ISME (for Incoherent Sampling of Multi-Echo data),
mitigates cardiac-induced noise along the echo dimension. Compared to standard linear sampling, ISME reduces the variability of R2* and QSM by respectively 21-26% and 23-32%, at
a cost of a scantimeincrease of 12%.