893 Views
Quick Fire Session
SCMR 22nd Annual Scientific Sessions
Olivier Jaubert
PhD Candidate
King's College London
Gastao Jose Lima Cruz, PhD
Postdoctoral Research Associate
King's College London
Aurelien Bustin, PhD
Research Associate
King's College London
Torben Schneider, PhD
Philips Clinical Scientist
Health Systems, Philips
Peter Koken, MSc
Scientist
Philips Research
Mariya Doneva, PhD
Senior Scientist
Philips Research
René Botnar, PhD, FSCMR
Chair of Cardiovascular Imaging
King's College London
Daniel Rueckert, PhD
Head of the Department of Computing
Imperial College London
Claudia Prieto, PhD
Reader
School of Biomedical Engineering and Imaging Sciences, King's College London
Background: Cardiac MRI is the gold standard to assess left ventricular function, whereas T1 and T2 mapping techniques are emerging to provide quantitative assessment of myocardial tissue properties. However, all these sequences are acquired sequentially under several breath-holds, thus resulting in long scan times and misaligned images. ECG-triggered cardiac MR fingerprinting (MRF)1,2 has been recently proposed to provide both T1 and T2 maps from a single scan at the diastolic cardiac phase. Cardiac MRF interrupts the acquisition and requires a subject-dependent dictionary to reconstruct the images. In this work, we aim to implement a motion-resolved cardiac MRF (MORE-MRF) acquisition and reconstruction for efficient assessment of cardiac tissue viability and function in a single breath-hold scan.
Methods: MORE-MRF employs a continuous bSSFP MRF acquisition with several inversion (IR) pulses (one IR every ~2.9s), varying flip angle pattern and tiny golden radial trajectory. Data is assigned to 8 cardiac phases using retrospective ECG gating. Time-point MRF images corresponding to each cardiac phase are reconstructed combining parallel imaging3, low rank approximation in the temporal direction4,5 and an extended local patch-based low rank regularization6. A dictionary of signal evolutions was computed using the hybrid state free precession framework7 for a range of T1 and T2 values. T1 and T2 maps were obtained for each cardiac phase through dot product matching between the reconstructed signal and dictionary. MORE-MRF acquisitions were performed in a standardised phantom (simulated heartrate of 65 bpm) and under breath-hold in 5 healthy subjects on a 1.5T scanner (Philips Ingenia). Imaging parameters include: TR/TE= 4/1.99 ms, 2x2x10 mm resolution, 256x256x10 mm FOV, 29s scan time. Additionally, conventional 2D CINE, MOLLI and SASHA T1 maps, and mGRASE T2 map were acquired for comparison purposes (with matching imaging parameters).
Results: Phantom results showed a maximum error under 6% in both T1 and T2 myocardium values when compared to gold standard spin echo scans. Cardiac-resolved T1 and T2 maps, together with the corresponding CINE sequence synthetically generated from the MRF maps are shown in Fig.1 for a representative healthy subject. The conventional CINE acquisition is also included for comparison purposes. MORE-MRF diastolic T1 and T2 maps are shown in Fig.2 in comparison to the gold standard clinical sequences for a healthy subject. Myocardial values measured in the septum for these sequences are shown in Table 1 for all subjects.
Conclusion: We propose a continuous MRF acquisition providing cardiac-resolved T1 and T2 maps for simultaneous assessment of cardiac tissue viability and cardiac function. Quantitative values are in general good agreement with those from SASHA and T2mGRASE, however confounding factors such as field inhomogeneities, magnetisation transfer and motion have been reported to affect MRF and will be investigated in future work.