2116 - Initial Experience with Using Deformable Registration Techniques to Calculate Prior Organ-at-Risk Dose for Re-Irradiation

Purpose/Objective(s): We report upon our initial experience with a protocol to rigorously account for prior organ dose for patients receiving radiotherapy re-treatment. We compare the calculated organ-at-risk (OAR) prior doses upon the re-planning CT using either rigid or deformable registration to transfer the prior dose.

Materials/Methods: For patients in our clinic who had received prior radiotherapy, we implemented a protocol to calculate the OAR dose from the prior treatments. An initial risk stratification was performed by converting the prior physical dose to 2 Gy per fraction equivalent dose (EQD2) assuming a/ß = 3 for all OARs, automatically rigidly registering the prior-treatment image to the re-treatment planning image, and mapping the prior EQD2 distribution to the re-plan CT. If for any OAR the prior EQD2_0.03cc value exceeds a tabulated organ-specific threshold, a high-risk EQD2 assessment is then manually performed. The high-risk assessment may use either a manual rigid registration or a muti-modality deformable registration algorithm between the prior-treatment and re-treatment CT images. The results from the high-risk assessment are used to inform the attending physician and to constrain the subsequent re-treatment plan as needed to ensure that cumulative OAR EQD2 tolerance will not be exceeded during re-treatment.

Results: We recorded 86 re-treated patient cases for which a high-risk assessment was required; 14 cases were re-treated in the abdomen, 35 in the chest, 19 in the head and neck, and 18 in the pelvis. Rigid registration was used in the high-risk assessment for 22% of all cases, however it was used for 58% of head and neck cases and for 12% of all other re-treatment sites. The greater use of rigid-only registrations for head and neck cases was driven primarily by previous physician experience. Among 15 head and neck cases (8 rigid and 7 deformable in the high-risk assessment), the difference in OAR dose between the automated risk stratification and the high-risk assessment was calculated for the brainstem, spinal cord, mandible, and larynx. The ratio of dose difference to risk-stratification OAR dose was in fact lower (p = 0.011) for deformable registration (11±13%) than for rigid registration (59±75%). OAR dose results from high-risk assessment using deformable registration appear to be more consistent with the initial automated risk stratification.

Conclusion: For re-treatment cases deemed to be high risk due to high prior OAR dose, we observed that rigid registration techniques were more likely to be used for head and neck cases than for other re-treatment sites. However, for head and neck cases using deformation for high-risk EQD2 assessment the OAR prior dose results are consistent with the initial automated risk stratification workflow; this gives confidence in the validity of using deformation methods for dose accumulation.