Characterization of a deformable beating cardiac phantom with real-time dosimetric capabilities for validation of...

Stereotactic arrhythmia radioablation (STAR) has emerged as a salvage treatment for patients with ventricular tachycardia, and the MR-linacs offer MRI-guidance during such treatment. However, available workflows on MR-linacs are not yet optimized and characterized dedicatedly to perform heart radiotherapy due to a lack of realistic MRI-compatible cardiac phantoms suitable for real-time dosimetry.

This work introduces a newly designed, deformable, and MRI-compatible cardiac phantom with real-time and multi-point dosimetric capabilities, characterizes its MRI properties and mechanical behavior, and showcases its use for MR-linac end-to-end workflow testing.

This cardiac phantom (IBA QUASAR, London, Ontario, Canada) is composed of a deformable heart model - representing the left and right ventricles - with integrated Plastic Scintillation Detectors (Medscint, Quebec, Quebec, Canada) (PSDs), a section filled with contrast solution, and a piston compatible with the motor of the QUASAR^TM MRI^4D motion phantom. We determined its T1 and T2 values by acquiring gold standard inversion recovery and spin echo sequences. We evaluated its 2D deformation and mechanical behavior by performing an MRI-based analysis inspired by dynamic mechanical analysis. We evaluated its 3D deformation and mechanical behavior by acquiring 3D scans at different motor positions, and by applying deformable image registration. We evaluated the reliability of the PSD measurements by testing the repeatability, linearity, and dose rate dependency of their measured dose on a 1.5 T Unity MR-linac (Elekta AB, Stockholm, Sweden). We evaluated the dosimetric impact of motion and motion mitigation on the dose measured by the PSDs by delivering a STAR plan in different cases: static, motion (cardiorespiratory or cardiac motion), and motion with gating (MR-linac clinical workflow).

On average, T1 values were T1_{constrast_solution} = 1130 $\pm$ 44 ms, T1_{left_ventricle's_wall} = 763 $\pm$ 76 ms, and T1_{right_ventricle's_wall} = 775 $\pm$ 69 ms, and T2 values were T2_{constrast_solution} = 155 $\pm$ 16 ms, T2_{left_ventricle's_wall} = 49 $\pm$ 12 ms, and T2_{right_ventricle's_wall} = 185 $\pm$ 35 ms. For the 2D deformation, no phase lag of deformation (i.e., ventricular area) with respect to the piston motion and no hysteresis behavior were observed. For the 2D and 3D deformation, linear relationships between piston positions and ventricular areas or volumes of interest were observed (R² = 0.99 and R² = 1.0, respectively). For the 3D deformation, negligible differences between volumes before and after PSDs integration were obtained (largest absolute percentage difference of 1.1%). The maximum deformation of the heart model was 5 mm. For the dosimetric capabilities, the reliability of the PSD measurements was shown (repeatability: coefficient of variation $\le$ 0.31%, linearity: R² = 1.0, dose rate dependency: coefficient of variation $\le$ 0.72%). Dose rates measured by the PSDs over time during STAR delivery were fluctuating in phase with the motion pattern. Percentage errors on cumulative doses with respect to static cases when cardiorespiratory or cardiac motion were applied were, respectively, up to -28.6% and -0.7% in motion cases, and up to -3.1% and 0.8% in gating cases.

This deformable and MRI-compatible cardiac phantom has good MRI contrast, and its heart model exhibits elastic behavior. Its integrated PSDs offer real-time and multi-point dose measurements, and their reliability was demonstrated. We successfully showed that this cardiac phantom enables end-to-end testing for heart radiotherapy on MR-linacs.