Purpose: Cardiac radioablation (CR) for atrial fibrillation (AF) presents one of the most challenging radiotherapy tracking scenarios: multiple small targets undergo rapid differentiational cardio-respiratory motion and are near organs-at-risk. MR-guidance offers a solution for real-time target tracking. This work aims to develop and investigate MRI-guided tracking for cardiac-induced motion of CR targets.
Methods: A tracking algorithm utilizing template-matching concepts, image manipulation optimized for target visualization and population-based target motion assumptions was developed. Tracking performance was characterized using breath-hold MRI images from six healthy and six AF participants. Multi-slice, ECG-gated, 2D cines were interpolated into 3D images for 18-30 cardiac phases, creating 4D datasets. Ground-truth targets were contoured on every cardiac phase. Retrospective tracking was performed on a single transverse slice of a 4D dataset. 3D geometric and volumetric tracking errors were calculated, defined as the difference between the ground-truth and tracked target centroids and volumes respectively. Single slice, transverse, ungated cines (50-60 images) were acquired at 4Hz in three healthy and three AF participants and used to characterize retrospective tracking performance on images more comparable to MRI-Linac acquisitions. Each image was visually assessed for critical tracking errors, defined as the tracked target overlapping an organ-at-risk.
Results: The average 3D geometric tracking error was 1.6±1mm and 1.3±1mm for healthy and AF participants respectively. 90% of 3D geometric tracking errors were less than 3mm and 71% were less than 2mm. The average volume difference for all evaluated images and targets was 0.27±3cc and 0.02±2cc in healthy and AF participants respectively. 8.0±5% and 11±10% of the ungated images in healthy and AF participants respectively had critical 2D tracking errors.
Conclusion: The feasibility of MRI-guided AF CR cardiac-induced target motion tracking has been demonstrated. The promising results warrant further investigation as a possible means to improve CR for AF treatment efficacy.
Funding Support, Disclosures, and Conflict of Interest: Funding: This project was funded by the Auckland Academic Health Alliance Collaboration Fund. Disclosure: The authors would like to thank MIM Maestro for use of a research license .