Purpose:Pencil beam dose calculation algorithm(PBDCA),due to its efficiency and robustness,has been widely used in most commercial treatment planning systems(TPS).The purpose of this study is to develop a PBDCA for a robotic radiosurgery system.Methods:A pencil beam dose calculation algorithm,which is widely used in Linac system,is adapted to a robotic radiosurgery system.The algorithm models both primary(short range)and scatter(long range)components with a single input parameter.The input parameter was first estimated to derive an initial set of pencil beam model parameters(PBMP).The agreement between predicted and measured TPRs were evaluated by the root mean square difference(RMSTPR),which was then minimized by adjusting PBMPs.PBMPs are further tuned to minimize the discrepancy of predicted and measured OCR by focusing at the outfield region.Finally,an arbitrary intensity profile is optimized by minimizing the OCR difference(RMSocr)at infield region.The developed PBDCA is tested on the CyberKnife system.We also compared with the ray-tracing method of CyberKnife system TPS on clinical treatment plans.Results:With RMS values normalized to the dmax and all cones combined,the average RMSTPR at build-up and descending region is 2.3%and 0.4%,respectively.The RMSocr at infield,penumbra and outfield region is 1.5%,0.6%and 0.87%,respectively.There is no trend found in TPR or OCR agreement among cones or depths.The comparison of clinical treatment plans demonstrate a better dose calculation accuracy than raytracing algorithm.Conclusion:A PBDCA for robotic radiosurgery system was developed.The prediction agrees well with commissioning data,and the clinical plans show better accuracy.Only a subset of measurements is needed to derive the model,which can be easily implemented in TPS.