365 - Simultaneous T1, T2, T2*, and PDFF Estimation Using Magnetic Resonance Fingerprinting on a 1.5T MR-Linac for Head and Neck Cancer Patients
Presenter(s)
L. McCullum1,2, S. Mulder1,2, E. D. Subashi3, K. P. Hwang4, C. D. Fuller2, and J. Ostenson5; 1UT MD Anderson Cancer Center UTHealth Houston Graduate School of Biomedical Sciences, Houston, TX, 2Department of Radiation Oncology, The University of Texas MD Anderson Cancer Center, Houston, TX, 3Department of Radiation Physics, The University of Texas MD Anderson Cancer Center, Houston, TX, 4Department of Imaging Physics, The University of Texas MD Anderson Cancer Center, Houston, TX, 5Department of Radiology, The University of Washington, Seattle, WA
Purpose/Objective(s): Recent work has shown that T1 intensity, T2 intensity, and proton density fat fraction (PDFF) change in salivary glands following radiation-induced damage. However, quantitative imaging suffers long scan times resulting in limited sequences available due to time constraints especially on the integrated 1.5T MRI / linear accelerator (MR-Linac). Therefore, simultaneous acquisition methods must be explored to fully take advantage of the high temporal density available on the MR-Linac. Magnetic resonance fingerprinting (MRF) is a recent innovation designed to simultaneously extract multiple quantitative property maps in a single, co-registered, acquisition. Therefore, the purpose of this study was to investigate the preliminary technical feasibility of MRF on the 1.5T MR-Linac to simultaneously quantify T1, T2, T2*, and PDFF.
Materials/Methods: Inversion-prepared multi-echo MRF data were acquired with a golden angle k-space trajectory on a 1.5T precision radiation medicine company Unity MR-Linac using TI = 100 ms, fixed TR = 20 ms, TEmin = 3.21 ms (?TE = 1.53 ms , 9 TEs), 2048 excitations with variable flip angle with RF-spoiling and high time-bandwidth product to reduce slice profile effects. The data were corrected for phase errors in the bipolar multi-echo acquisition followed by low rank reconstruction with total variation regularization of each TE with a pre-calculated signal-subspace basis from a dictionary generated from an extended phase graph model of the MRF sequence with log-spaced T1 / T2 scaling. The first singular image were then jointly fit for off-resonance, phase offset, T2*, and an approximate PDFF map using Phase Regularized Estimation using Smoothing and Constrained Optimization (PRESCO). The complete set of B0/T2*-corrected fat-water separated water singular images were then estimated using the maps from PRESCO, and then fitted for T1 and T2 using dictionary matching. The ISMRM/NIST Phantom was used to compare measurements with T1 and T2 reference values. Independent B0 and B1+ maps were also acquired for corrections. A healthy volunteer was scanned with IRB approval and informed consent to validate T2* and PDFF performance.
Results: Lin's concordance correlation coefficient (LCCC) was >0.9 in the ISMRM/NIST phantom for both T1 and T2 compared to reference values. Further, mean bias was within 6%. T2* and PDFF values agreed with expected literature values for all structures analyzed.
Conclusion: MRF has demonstrated the potential to replace the current time-intensive sequences on the 1.5T MR-Linac, thus increasing the adoption potential of more specialized quantitative imaging biomarkers approaches at high temporal density.
Abstract 365 - Table 1| T1 (ms) | T2 (ms) | |
| ISMRM/NIST Phantom T1 Vials | LCCC = 0.91 Mean Bias = 5.38% | LCCC = 0.93 Mean Bias = -1.51% |
| Median T2* (ms) | PDFF (%) | |
| Left Parotid | 19 | 53 |
| Right Parotid | 26 | 51 |
| Left Masseter Muscle | 24 | 1 |
| Right Masseter Muscle | 23 | 1 |
| Posterior Neck Fat | 71 | 98 |