Ziegenhein Peter, Kamerling Cornelis Ph, Fast Martin F, Oelfke Uwe
Joint Department of Physics at The Institute of Cancer Research and The Royal Marsden NHS Foundation Trust, London, SM2 5NG, UK.
Sci Rep. 2018 Feb 26;8(1):3662. doi: 10.1038/s41598-018-21966-x.
In this work we describe an ultra-fast, low-latency implementation of the energy/mass transfer (EMT) mapping method to accumulate dose on deforming geometries such as lung using the central processing unit (CPU). It enables the computation of the actually delivered dose for intensity-modulated radiation therapy on 4D image data in real-time at 25 Hz. In order to accumulate the delivered dose onto a reference phase a pre-calculated deformable vector field is used. The aim of this study is to present an online dose accumulation technique that can be carried out in less than 40 ms to accommodate the machine log update rate of our research linac. Three speed optimisation strategies for the CPU are discussed: single-core optimisation, parallelisation for multiple cores and vectorisation. The single-core implementation accumulates dose in about 1.1 s on a typical high resolution grid for a lung stereotactic body radiation therapy case. Adding parallelisation decreased the runtime to about 50 ms while adding vectorisation satisfied our real-time constraint by further reducing the dose accumulation time to 15 ms without compromising on resolution or accuracy. The presented method allows real-time dose accumulation on deforming patient geometries and has the potential to enable online dose evaluation and re-planning scenarios.
在这项工作中,我们描述了一种超快速、低延迟的能量/质量转移(EMT)映射方法的实现,该方法利用中央处理器(CPU)在诸如肺部等变形几何体上累积剂量。它能够以25Hz的频率实时计算调强放射治疗在4D图像数据上实际交付的剂量。为了将交付的剂量累积到参考相位上,使用了预先计算的可变形矢量场。本研究的目的是提出一种在线剂量累积技术,该技术可以在不到40毫秒的时间内完成,以适应我们研究直线加速器的机器日志更新速率。讨论了针对CPU的三种速度优化策略:单核优化、多核并行化和向量化。在一个典型的高分辨率网格上,对于肺部立体定向体部放射治疗病例,单核实现累积剂量大约需要1.1秒。添加并行化将运行时间减少到大约50毫秒,而添加向量化通过进一步将剂量累积时间减少到15毫秒,在不影响分辨率或准确性的情况下满足了我们的实时约束。所提出的方法允许在变形的患者几何体上进行实时剂量累积,并有可能实现在线剂量评估和重新规划方案。