Implementation of a Monte Carlo radiation transport code on Graphics Processing Units (GPU) for accelerated dose calculations in radiation oncology

在放射治疗中，基于计算机的计算是必不可少的，以确定将辐射输送到疾病部位的最佳方式。 这些计算的准确性传统上受到临床约束的限制，这些临床约束要求计算时间短。因此，执行这些计算的最准确的方法-蒙特卡罗模拟-在临床中没有广泛使用，因为在标准PC上的执行时间相对较长。虽然存在蒙特卡罗模拟的快速替代方案，但它们仍然构成了近似的解决方案。此外，这些技术几乎不能用于磁场中的计算，这是目前正在开发的创新系统中所需要的，并且将放射治疗设备和磁共振扫描仪组合用于精确的患者定位（MRI-直线加速器）。特别是对于MRI直线加速器和一般的放射治疗，快速蒙特卡罗计算引擎是长期追求的目标。该研究项目与Elekta Canada和Elekta CMS Software合作进行，旨在通过使用配备高端显卡的PC将高性能Monte Carlo计算带入临床。这些图形卡可以执行并行计算，最近被证明在蒙特卡洛模拟的执行速度方面取得了惊人的进步。他们设想在临床使用的治疗计划系统的重大改进。快速蒙特卡罗代码也构成了MRI直线加速器发展的基石，这是一种有前途的放射治疗技术。

Computer-based calculations are essential in radiotherapy to determine the best way to deliver radiation to the disease site. The accuracy of these calculations has traditionally been limited by clinical constraints that dictate a short calculation time. Consequently, the most accurate method to perform these calculations -- Monte Carlo simulations -- is not widely used in the clinic because of relatively long execution times on standard PCs. While fast alternatives to Monte Carlo simulations exist, they nonetheless constitute an approximated solution. Furthermore, these techniques can hardly be used for calculations in a magnetic field, which are required in innovative systems currently in development and combining a radiotherapy device and a magnetic resonance scanner for accurate patient positioning (MRI-linac). For a MRI-linac in particular and for radiotherapy in general, a fast Monte Carlo calculation engine is a long sought-after objective. This research project, conducted in collaboration with Elekta Canada and Elekta CMS Software, is aimed at bringing high-performance Monte Carlo computing to the clinic through the use PCs equipped with high-end graphics cards. These graphics cards can perform parallel calculations and were recently shown to achieve spectacular improvements in the execution speed of Monte Carlo simulations. They let envision major improvements in treatment planning systems used in the clinic. A fast Monte Carlo code also constitute a cornerstone in the development of a MRI-linac, a promising technology in radiation therapy.