Abstract: Objective To investigate the feasibility of
applying a RapidPlan model trained with pre-surgical rectal cancer
simultaneous integrated boost (SIB) plans to the optimization of
post-surgical non-SIB plans, and to explore the methods to improve the
existing models and extend their applications. Methods An existing
RapidPlan model for pre-surgical rectal cancer SIB plans (PTV/PGTV: 41.8
Gy/50.6 Gy) was modified and used to re-optimize 18 clinical
post-surgical non-SIB plans among which 7 and 11 cases were prescribed
with 45 Gy and 50 Gy to PTV, respectively, without changing the original
beam geometries, energies, accelerator configuration and dose volume
calculation algorithm. The target structures were matched with PTV and
PGTV in the model. The realizable DVH intervals were predicted, and the
re-optimized plans were renormalized to achieve similar target coverage
of the clinical plans before dosimetric comparison. Results Severe hot
spots were caused by matching target areas of non-SIB plans to PTV in
the model, while re-optimized plans could achieve similar or better plan
quality with smaller standard errors of organs-at-risk dosimetric
parameters, compared with the clinical plans. Conclusion RapidPlan model
for pre-surgical rectal cancer SIB plans can be applied to the
automatic optimization of post-surgical non-SIB plans with higher
consistency in plan quality, but the target structure is needed to be
matched to the high prescription target structure in SIB model.