Knowledge-based (KB) plan prediction model for prostate SBRT delivered with CyberKnife according to the RTOG0938 protocol.

The Aim is to extend the KB approach using the RapidPlan (RP) tool to SBRT with CyberKnife (CK) in the case of prostate treatment. Seventy-two prostate clinical plans of patients treated at our institute with CK between 2017 and 2021 were exported from CK system (Accuray Precision v.3.3) to Eclipse (Varian v.13.6). All patients were treated according to the RTOG 0938 protocol, delivering $36.25 {Gy}$ in 5 fractions. Bladder, rectum and femoral heads were considered in the KB model. Fifty-two randomly chosen plans were used to validate it. An individually tailored CK template for automatic plan optimization was created. All automatic plans (KB-TP) were compared against the original clinical plans (TP) in terms of OARs/PTVs dose-volume parameters. OARs sparing for KB-TP plans were generally better than TP plans, with a slightly worsening of PTV coverage. The feasibility of building a robust KB model for SBRT application with CyberKnife was demonstrated.