ESTRO 2024 - Abstract Book

S4033

Physics - Inter-fraction motion management and offline adaptive radiotherapy

ESTRO 2024

The ANACONDA algorithm (RayStation 11A) was used for all deformations. All OAR structures were outlined by a single consultant oncologist, with the following OARs evaluated: brainstem, parotid, larynx, spinal canal, optic chiasm, orbits, optic nerves, and mandible OARs for both CT’s.

Results:

Figure 2 summarises results for the novel RMOD approach as well as comparator methods.

The rigid (#1) registrations showed the largest difference between original and mapped OAR near-maximum doses (D0.1cc). By using image intensity to drive a deformable (#2) registration there was a marked reduction in dose differences, except for optic chiasm and nerves. Per organ deformations (#3) using single OAR contours reduced differences in doses considerably, however there were still a number of positive and negative outliers. Using the global hybrid (#4) deformation improved agreement relative to (#2), however conflict between deformation of different OARs limited the improvements relative to (#3). Finally, the RMOD approach (#5) demonstrated a significant reduction in low dose differences, although still with some spread of dose present. Overall, RMOD was least prone to dose underestimation across all OAR; any remaining dose underestimation correlated with the most challenging changes in patient anatomy. Additionally, due to the resampling process, the geometric spread of point maximum doses was increased by the robust approach, which should improve confidence that doses are not being under reported. RMOD was however associated with 0.1 to 3.3% higher doses than values based on the original CT.