S1696
Physics - Detectors, dose measurement and phantoms
ESTRO 2026
Deformable Image Registration (DIR). We propose a phantom-based verification of dose accumulation for motion consideration to demonstrate impact of DIR uncertainty on calculated dose. Material/Methods: A simple cork-solid water slab phantom (Figure 1) was used digitally to avoid possible imaging artifacts. Eight phases/CTs representing combinations of four respiratory (from full-expiration to full-inspiration of 3cm amplitude) and two cardiac phases (systole/diastole of 1cm amplitude) were acquired (Figure 2). CTs were rigidly registered using fiducial markers to simulate target tracking dose delivery. Dose accumulation using 4 different approaches was performed, including DIR for motion consideration. Calculations were based on the mean doses in a Test Volume (TV) located 1cm below ‘heart’ anterior surface.4D/FRAGMENTED/NO-DIR: cardio-weighted sum of respiration-based fragmented plan’s doses calculated on each of 4D-CTs for the TV. This approach was set as reference because it depends only on respiratory motion uncertainty. 4D/FRAGMENTED/DIR: same as method-1, but TV doses obtained by dose warping complementary CT phases to full expiration in diastole phase using DIR.Semi4D/FULL/NO-DIR: cardio- resp-weighted sum of the original plan recalculated doses on each of complementary 4D-CTs for the TV.Semi4D/FULL/DIR: same as method-3, but TV doses obtained as in method-2.Dose calculations were done using in-house MATLAB scripts [2]. DIR and dose warping were performed with 3D-Slicer, RayStation, and MIM.
Results: Results based on the %deviation from the reference (4D/FRAGMENTED/NO-DIR):
Anatomical variation produced dosimetric deviation ≈4-5% under simulated phantom conditions. 4D/FRAGMENTED and Semi4D/FULL approaches were found equivalent. Guided (by using controlling volumes) DIR from RayStation approach produced consistent dose error ≈0.5%, while unguided DIR from 3D-Slicer and MIM ≈2%. Conclusion: Neglecting anatomy variation and deformation in standard-CK planning can be associated with 4-5% error, potentially relevant in high-dose STAR. This error can be eliminated using (Semi)4D approaches for dose calculation, but control of methods’ uncertainty is crucial. Ongoing work focuses on validation with Probability Distribution Function method [3], and then implementation on real patient
data.Acknowledgments: CTU has a scientific collaboration with RaySearch Laboratories. References:
Made with FlippingBook - Share PDF online