Purpose: We present FRED, a GPU-accelerated monte carlo (MC) tool for proton therapy dose calculation and research. We developed an automated FRED commissioning and validation procedure exploiting proton facility commissioning data. We give examples of FRED clinical applications, i.e. physical and radiobiological dose calculations performed on proton therapy patients.
Methods: We developed python software tools and used facility commissioning measurement to implement phase space library in FRED with minimal manual intervention. We validated FRED dose calculations against measurements of: (i) SOBPs of various ranges and (ii) patient QA plans in a water phantom with ionization chamber. To validate the CT scanner calibration, we measured monoenergetic fields with 2D array of ionisation chambers (Matrix from PTW) positioned behind a heterogeneous CIRS phantom. We used FRED to recalculate treatment plans of patients treated in our proton beam therapy facility. We evaluate the results by comparing dose distribution parameters and via the gamma index (GI) method.
Results: FRED offers a proton tracking rate of 10^7 protons/s using two GPU cards (Nvidia Titan X). We obtained submillimeter agreement of FRED pencil beam dose calculations compared with measurements (beam range, Bragg peak FWHM, sigma of lateral profile). The SOBP dose calculations agree with measurements within 2%. The comparison of 182 patient transversal field computations and Matrix measurements gives GI=97.9(3.3)% (3%/2 mm, 1 sigma). The GI pass rate when comparing FRED calculation and Matrix measurement of a two dimensional dose distribution behind the CIRS phantom irradiated with a monoenergetic field is better than 99%.
Conclusion: These results confirm excellent performance of the physics models implemented in FRED. The presented results show that the FRED dosimetric accuracy enables its application in clinical routine and potential improvement of patient treatment with protons.