AbstractOver the past few decades Magnetic Resonance Imaging (MRI) has become a clinically important medical imaging modality, thanks to its superior soft tissue contrast over computed tomography (CT), and the beneﬁt of imaging deep anatomies which is challenging with ultrasound technology. In addition, unlike CT and PET it does not make use of ionizing radiation. This makes MRI well-suited for at-risk patient populations, as well as an indispensable modality for healthcare research.
Unfortunately MRI is also expensive, largely due to the cost of hardware involved and its maintenance. Adding to this problem, MR acquisition times are typically longer than comparable CT examinations, which can make MRI scans uncomfortable experiences, and economically speaking decreases patient throughput. Both of these disadvantages are addressed by reducing MR scan durations, where parallel imaging methods such as SENSE, GRAPPA and compressed sensing have had success in the past decade. Acceleration in parallel imaging however comes at a cost in signal-to-noise ratio (SNR) as MR images are formed from undersampled data.
Multiband techniques (a.k.a. Simultaneous Multi-Slice) reduce scan time by exciting and acquiring signal from multiple slices simultaneously, and consequently using multiple receiver coils for unfolded reconstruction. This type of acceleration has an SNR beneﬁt over parallel imaging, for a ﬁxed echo-time, as when no in-plane undersampling is used the SNR cost is solely due to multi-channel reconstruction. However challenges in multiband imaging arise in the design of multiband RF pulses, which can take signiﬁcantly longer to transmit and consequently lead to lower acquired signal, due to relaxation. Exciting multiple slices also places signiﬁcant demand on current hardware, in terms of coping with higher RF power, and higher frequency demands for both RF and gradient systems. Up-grading hardware is not always an option, due to the cost involved. In addition, higher RF power requirements for exciting multiple slices can also lead to increased patient heating, measured as speciﬁc absorption rate (SAR), which needs to be considered for MRI safety.
The work presented in this thesis seeks to improve multiband RF pulse design techniques by three means; ﬁrstly, we demonstrate practical hardware considerations for implementing time-optimized multiband pulses. This includes considerations for RF transmission hardware for phase-optimized, time-shifted and root-ﬂipped multiband RF pulses, as well as gradient hardware considerations for using time-variable selection gradients, which have been shown to greatly enhance multiband RF performance. Based on such considerations, we propose the use of amplitude modulated multiband RF pulses, as well as the design of time-variable gradient waveforms with smooth shapes, which signiﬁcantly reduce demands on MR hardware when necessary.
Secondly, time-shifted and root-ﬂipped multiband pulses can have misaligned spin-echo
signals for diﬀerent slices, which can lead to diﬀerent T2 and T2∗ contrast; these eﬀects have been investigated and discussed.
Thirdly, using the hardware considerations we have implemented multiband RF pulses with time-variable selection gradients in a cardiac bSSFP sequence, one of the most clinically relevant examinations of our time. Although the shortest RF pulses are thought to be optimal for rapid MR, they do not directly lead to the shortest acquisition time due to patient heating, as this is limited by SAR. We reformulated the RF pulse design problem to reduce image acquisition duration directly which can result in shorter breath-hold periods and improved image quality.
|Date of Award||2019|
|Supervisor||Shaihan Malik (Supervisor) & Jo Hajnal (Supervisor)|