Room: Exhibit Hall | Forum 3
Purpose: ARCHER is a new, commercialized, Monte Carlo based radiation dose computing software package that leverages NVIDIA GPUs to significantly reduce computational time yet retain the dose calculation accuracy of the Monte Carlo method to busy and resource-limited oncology clinics. This work presents the heterogeneous computing design present in ARCHER, clinical workflow modules (PACS, DICOM import/export, GUI), and the benchmarking of a patient specific, IMRT capable, model of a Varian TrueBeam STx 6X accelerator in ARCHER using the well-benchmarked EGSnrc code package.
Methods: In ARCHER, an asynchronous global scheduler was designed to distribute tasks between source term generation and Monte Carlo particle transport within the patient to accelerate the computation time. Validation of ARCHER was completed by validating the EGSnrc model against a clinical treatment planning system model; ARCHER was then validated against the EGSnrc model. Multiple benchmarking scenarios included percent depth dose and axial profiles, MLC test patterns and one IMRT patient plan. Gamma analyses were conducted to establish the accuracy and consistency between the two Monte Carlo models.
Results: ARCHER is dramatically faster than EGSnrc yet retains computational accuracy. There were no dose differences in excess of 1% for percent depth dose and axial profiles. For the MLC test pattern, the 3%/3mm gamma pass rate between EGSnrc and Pinnacle was in excess of 99% and between EGSnrc and ARCHER the pass rate was 98.6%. Preliminary results of a clinical IMRT breast case are presented.
Conclusion: A patient specific, IMRT capable, model of the Varian TrueBeam STx 6X has been successfully developed in EGSnrc and has been used to successfully validate a similar model implemented with a heterogeneous computing design in ARCHER. This work demonstrates that Monte Carlo based dose calculation algorithms have the potential for widespread clinical implementation.
Funding Support, Disclosures, and Conflict of Interest: Grant support was provided from the National Institute of Biomedical Imaging and Bioengineering (STTR 4R42EB019265)