A new approach to off-line setup corrections: combining safety with minimum workload

Off-line patient setup correction protocols based on electronic portal images are an effective tool to reduce systematic patient setup errors. Recently, we have introduced the no action level (NAL) protocol which establishes a significant error reduction at a very small workload. However, this protocol did not include an explicit verification of the applied setup corrections. Systematic mistakes in the execution of setup corrections (e.g., a setup correction is always executed in the +X direction whereas a correction in the -X direction was prescribed) may introduce large systematic setup errors (irrespective of the setup protocol) and may seriously impair treatment outcome. We have therefore extended the NAL protocol with a correction verification (COVER) stage, solely aimed at detecting such mistakes. In short, COVER tests the magnitude of the postcorrection setup error in each relevant direction. If these residue errors are below the acceptance threshold T, no more electronic portal images are required and the protocol has finished. If not, the origin of this result should be investigated; if no obvious mistakes are present, the procedure is repeated for one more treatment fraction. If the residue setup errors are confirmed to be larger than T, the entire protocol is restarted. Using both Monte Carlo simulations and analytical calculations, we performed a risk analysis and evaluated the workload for various choices of T. A threshold T = 3 x sigma(r), where sigma(r) is the mean standard deviation of the random setup errors, ensured that (1) COVER introduces only a small additional workload (1.05 measurement per patient, while the absolute minimum is 1.0) and (2) serious correction mistakes are detected with high probability. Even if setup corrections are wrongly applied in each patient (worst case scenario), COVER ensures that the final distribution of systematic errors is not wider than the precorrection distribution of systematic errors; for realistic frequencies of correction mistakes (<< 1 per patient) this distribution becomes much more narrow. The combination of NAL and COVER thus provides a highly efficient as well as safe method to reduce systematic setup errors.