Evaluating CBCT-Guided Adaptive Radiotherapy for Pancreatic Cancer Using Synthetic CBCT Data

Abstract

Ethos adaptive radiotherapy is employed frequently in the pelvis to improve treatment accuracy by adapting to daily anatomical changes. The use of this CBCT-guided platform for abdominal treatments is made challenging by motion-related image artifacts that are detrimental to the Ethos auto-contouring process. We present a preliminary in silico study enabled by synthetic CBCT data of Ethos adaptive radiotherapy for pancreatic cancer. Simulation CT and daily CBCT images were collected from nonadaptive patients treated on Ethos. Contoured CBCTs drove structure-guided deformable registration from the CT to daily CBCTs, providing an approximate daily CT used to produce synthetic CBCT data. Two adaptive workflows were simulated using an Ethos emulator. Over 70 fractions across 10 patients in a solely deformation-based workflow, PTV prescription coverage increased by 23.3±9.4% through plan adaptation. Point doses to the stomach were reduced by 10.2±9.3%. Ultimately, un-adapted plans satisfied target coverage and OAR constraints in 0% and 6% of fractions while adapted plans did so in 80% of fractions. Anatomical variation led to poor performance in rigidly aligned un-adapted plans, illustrating the promise of Ethos adaptive radiotherapy in this region. This promise is balanced by the need for artifact reduction and questions regarding auto-contouring performance in the abdomen.