Purpose: The current MR-guided-radiotherapy(MRgRT) motion tracking was mostly based on the limited number of representative 2D slices while lacking volumetric information due to the insufficient MRI acquisition speed. We propose a real-time motion tracking solution for MRgRT using an ultra-fast volumetric 4D-MRI generated VOI positional probability distribution(VOI-PPD).
Methods: A free-breathing volumetric 4D-MRI (CAIPIRINHA-accelerated 3D gradient-echo-sequence with1.63 volume/sec, 56 axial-slice/volume, voxel size=2.7x2.7x4mmÂ³, 144 dynamics) was developed and applied on 9 healthy subjects (34.33Â±5.77 years) using a 1.5T MR-simulator. Left kidney(LK) and right kidney(RK) motion tracking were retrospectively conducted. Time-resolved motion was expressed as a transformation matrix (Tvoi) by linearly registering the dynamic target VOI to a reference volume such as planning GTV. A binary VOI mask was created on the reference volume and transformed by Tvoi for the generation of dynamic mask (M(j), j: time-frame index). The dynamic VOI-PPD at overlapping probability â‰¥ i%(Váµ¢) was estimated and updated at each time-frame by calculating the overlapping volume, (M(1)+...+M(j))/j. A baseline VOI-PPD was calculated using the first 10 time-frames. The dynamic VOI-PPD with positional probabilityâ‰§i%(Vi%) was then updated at every time frame. To evaluate the accuracy of the VOI-PDD approach, the V10%at the last time-frame 144 was also compared to the MIP volume at 40-60% respiratory phase(VÊ³áµ‰Ë¢áµ–â‚„â‚€â‚‹â‚†â‚€) using signed-rank test.
Results: The dynamic V10% stabilized after the first 10 time-frame (i.e.~10s). For both kidneys, similar Vâ‚?â‚€ were observed at time-frame of 44 (LK=190Â±36mmÂ³, RK=181Â±33mmÂ³), 64 (LK=192Â±35mmÂ³, RK=181Â±33mmÂ³),84 (LK=191Â±35mmÂ³, RK=181Â±33mmÂ³), 104 (LK=193Â±35mmÂ³, RK=180Â±32mmÂ³) and 144 (LK=192Â±35mmÂ³, RK=180Â±33mmÂ³). Significantly smaller Vâ‚?â‚€ than VÊ³áµ‰Ë¢áµ–â‚„â‚€â‚‹â‚†â‚€ (Vâ‚?â‚€ vs VÊ³áµ‰Ë¢áµ–â‚„â‚€â‚‹â‚†â‚€: LK=192Â±35 vs 207Â±35cmÂ³, p<0.05; RK=180Â±33 vs 190Â±29cmÂ³, p<0.05) was observed in both kidneys, indicating a more accurate motion tracking using the proposed VOI-PPD approach.
Conclusion: A motion tracking approach using an ultra-fast volumetric 4D-MRI generated VOI-PPD was proposed and might benefit motion tracking accuracy for real-time MRgRT.