2368 - Feasibility of Longitudinal Relaxation Rate Mapping with Non-Cartesian Sequences on a 1.5T MR-Linac

L. McCullum¹, K. P. Hwang², Y. Ding³, C. Tang⁴, J. Yang³, P. Balter³, J. Wang³, C. D. Fuller¹, and E. D. Subashi³; ¹Department of Radiation Oncology, The University of Texas MD Anderson Cancer Center, Houston, TX, ²Department of Imaging Physics, 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 Genitourinary Radiation Oncology, Division of Radiation Oncology, The University of Texas MD Anderson Cancer Center, Houston, TX

Purpose/Objective(s): The integrated MR-Linac (MRL) provides a novel platform for the delivery of precision MR-guided adaptive radiotherapy. While functional and metabolic imaging biomarkers are known to be predictors of malignancy and treatment response, the inclusion of these datasets in workflows for adaptive planning remains under investigation. Quantitative mapping of the longitudinal relaxation rate (R1) is a major building block for several multiparametric MRI protocols. The implementation of these protocols is challenging in anatomical sites that experience large physiological motion. We investigate the feasibility of using non-cartesian k-space sampling for longitudinal relaxation rate mapping on a 1.5T MRL. Materials/

Methods: All experiments were performed on an integrated 1.5T MRL. R1-mapping was implemented with the spoiled steady-state gradient-echo sequence and variable flip-angle (VFA) method. Bias and reproducibility were assessed in the NIST phantom and healthy volunteers. Two non-Cartesian sampling strategies were investigated: pseudo-randomized 3D radial and 3D stack-of-stars (SoS), each with half or full-spoke coverage. A Cartesian acquisition was included for benchmarking. The imaging parameters were: TR/TE=10/0.2ms and 10/1.7ms for half and full spoke, FA={3°, 20°}, FOV=25x25x25cm³, acquisition voxel=2x2x2mm³. Undersampling was varied by a combination of total number of spokes and parallel imaging. Total scan time was 2-5 mins. Repeat scans were acquired over five days to evaluate longitudinal bias and reproducibility using the coefficient of variation (CoV). Spearman’s test was used to evaluate the correlation of bias with measured R1.
Results: In the phantom, bias and reproducibility were comparable across sequences, shown in Table 1. Median bias was less than 3.7% with respect to either nominal or Cartesian values while reproducibility was within 9.0%. Short term bias and reproducibility over five days was within 3.1% and 4.4% respectively. For half-spoke acquisitions, the bias was not significantly correlated with R1, but there was a tendency towards significance in full-spoke acquisitions. In volunteer scanning, R1 values in paravertebral muscle, kidney cortex/medulla, prostate, and parotids reveal a large bias and uncertainty in structures affected by motion artifacts.
Conclusion: This work implements and evaluates clinically feasible R1-mapping methods with non-Cartesian sampling on a 1.5T MRL. The methods complement current protocols of quantitative MRI in anatomical regions with large physiological motion.

	Median Bias / CoV vs. Reference (%)	Median Bias / CoV vs. Cartesian (%)	Longitudinal Bias / CoV vs. Reference (%)	Spearman’s Test for Bias vs. Reference Across R1 (p-value)
Half-Spoke Radial	0.6 / 8.8	1.7 / 9.0	2.8 / 3.0	0.8
Full-Spoke Radial	2.6 / 8.9	3.7 / 9.0	2.6 / 3.7	0.08
Half-Spoke SoS	-0.7 / 8.1	3.6 / 8.1	-1.2 / 4.4	0.2
Full-Spoke SoS	3.4 / 8.0	1.4 / 8.0	3.1 / 1.7	0.01