3D motion validation with clinically used cine-MRI and a MR-LINAC phantom

Purpose/Objective:

Magnetic resonance (MR)-guided radiotherapy shows high potential to improve the precision and accuracy of radiation therapy. Especially hybrid MR-LINAC devices provide the possibility to perform on-line MR imaging during dose delivery and offer efficient tumour tracking or gating techniques for high-precision treatment of moving tumours. For the commissioning of such systems, the accuracy and reliability of real-time motion tracking through MR-imaging needs to be assessed. In this study we evaluate the dynamic target localization accuracy of a programmable MRI-compatible motion phantom using clinical cine-MRI sequences in all three spatial dimensions.
Material/Methods:
The phantom (CIRS Model 008M MRI-LINAC Dynamic Phantom) has a body representing a human thorax in shape and proportion that was filled with a 6,61g/l NaCl water solution. It incorporates an off-centric cylindrical rod with embedded gel-based target that can be moved and rotated through a programmable actuator. Sinusoidal motion trajectories with patient-oriented breathing frequencies (0.1-0.2 Hz) and motion amplitudes (5 mm – 20 mm) in all three spatial dimensions were programmed in the phantom’s Motion Control Software. Balanced steady-state free procession sequences (TE/TR=2.3/4.6 ms; FOV=300×300×150 mm³; Res=1.34х1.34 mm², SliceThickness=7 mm, FA=30°) were acquired in cine mode on a 3T MR scanner (Philips Achieva) with a time resolution of 489 ms. The center-of-mass motion of the target was extracted from the cine images using a manual segmentation-based procedure. Both the measured frequency and amplitude were compared to the programmed motion parameters. The frequencies were determined with a Fast Fourier Transform (FFT). The phantom was also fed with a real patient’s 1D-navigator-based breathing pattern to evaluate the accuracy of non-regular target motion detection.
Results:
The frequencies (f) and amplitudes (A) extracted from the cine-MRI are in good agreement to the pre-set values. For the sinusoidal motion patterns, we observed 2% deviations between the measured and pre-set frequencies in IS direction for f=0.1 Hz and f=0.2 Hz with A=20 mm. In AP and LR direction the frequency deviation is 3% for f=0.2 Hz and A=5 mm. The amplitudes were determined with a precision of 99% in IS, and 92% in AP/LR direction with deviations smaller than 0.4 mm. For the real patient’s navigator breathing-pattern with main frequency components between 0.14-0.2 Hz and amplitudes between 5-20 mm we observed an amplitude accuracy of 98% with a maximum deviation of 1.2 mm in IS direction. The uncertainties in frequency and amplitude are dominated by the spatial and time resolution.
Conclusion:
The study shows motion parameters of the MRI-LINAC Phantom to be extracted from cine-MRI with high accuracy. Dynamic target localization through cine-MRI is feasible and accurate for the management of respiratory motion in radiation oncology.