Investigation on Performance Accuracy of Different Surrogates in Real Time Tumor Tracking at External Beam Radiotherapy

In external beam radiotherapy of dynamic tumors, several errors raise due tointer- and intra-fractional motions. In order to compensate these errors, signals obtainedfrom different surrogates are used to infer with tumor motion as real time. Therefore,a comparative assessment may be worthwhile on the effect ofdifferent surrogates intumor motion tracking. The performance accuracy of three internal-external surrogates entitled:external markers, diaphragm movement and lung volume was done using 4Dimensional Extended Cardiac-Torso (4D-XCAT) phantoms. Adaptive NeuroFuzzy Inference System (ANFIS) model was implemented to correlate the motionof surrogates with several tumors located in liver and lung, separately. Finally,the Root Mean Square Error (RMSE) of ANFIS model outputs in tumor motionpredictionof different surrogates was compared as metric tool. The average value of RMSE of lung and liver tumors were 0.4 mm, 0.6 mmand 0.8 mm for external markers, lung volume and diaphragm motion, respectively. Among three investigated surrogates, the best performance belonged toexternal markers strategy, while optimum location of these markers determined usingan input selection algorithm in this method.