Background: Cardiac magnetic resonance imaging (CMR) is widely regarded as the gold-standard imaging modality for evaluating cardiac structure, function, and viability. However, a significant rate of non-diagnostic scans (up to 25-50%) is observed in patients with cardiac implanted electronic devices (CIED) due to susceptibility and off-resonance artefacts caused by the device. Every year in the UK, 60,000 CIEDs are implanted and CMR in these patients remains an important clinical challenge. New MRI scanners operating at much lower magnetic field (0.55T, instead of 1.5T/3T currently in clinical use) which have much lower cost, reduced susceptibility artefact from CIED and further improved safety and thus represent a promising opportunity for imaging of CIED patients.
Aims: The proposed research aims to develop a novel CMR protocol for CIED patients:
Aim 1/Year 1: Development of robust off-resonance insensitive RF pulses (wideband pulses) for low field CMR to enable basic acquisitions such as late gadolinium enhancement (scar imaging), and quantitative T1/T2 mapping. Automated optimisation of these pulses using calibration scans and deep learning will be explored.
Aim 2/Year 2: Development of accelerated acquisition and reconstruction techniques to enable fast protocol and to minimise the time spent by the patient inside the scanner. The use of deep learning will be explored for fast reconstruction of highly accelerated acquisitions such as simultaneous multi slice imaging.
Aim 3/Year 3: Clinical evaluation of these techniques in CIED patients.
Learning experience: The student will develop a range of inter-disciplinary skills including in cardiology, AI/deep learning, MRI physics, Medical imaging acquisition/reconstruction, computer programming.
Figure 1. Current limitations of CMR techniques. A) Current clinical images in patients at 1.5T/3T which are associated with 25-50% rate of non diagnostic scans. B) highly accelerated scans at 1.5T using simultaneous multislice imaging which currently requires very long reconstruction times (several hours) incompatible with a clinical workflow. C) Fast deep learning reconstruction of standard acquisition (non multi-slice) at 1.5T which remains associated with image blurring and insufficient image quality.
