目的 流血效果是虚拟手术模拟器视觉效果的重要组成部分，血流与固体交互的庞大计算量使取得实时的流血模拟效果具有很大的挑战性。提出一种基于图形处理单元(GPU)加速的虚拟手术流血效果模拟方法。方法 该方法以Müller等人提出的光滑粒子动力学(SPH)作为基础，采用温度项使粒子具有不同速度模拟血流形成的血槽，同时基于构建均匀空间网格的思想，利用通用并行计算架构(CUDA)多线程并行加速技术完成粒子控制方程的求解和血流与固体交互的计算，从而取得实时的效果。结果 实验结果表明，本文方法能够满足虚拟手术中切割表面流血和血液在器官中流动的模拟需求，在粒子个数为9000时仅需20 ms，对比于纯CPU的实现取得20.15倍的加速比，实现了大量粒子下的实时流血模拟。 结论 本文方法具有较好的灵活性和实时性的特点，可以应用于虚拟手术仿真系统之中。

Objective Bleeding simulation is an important visual effect. Real-time bleeding simulation is challenging because of the extensive computations required for blood-solid interactions. This study proposes a simulation method for surgical bleeding based on graphics-processing units (GPUs). Method The proposed method is derived from the smooth particle hydrodynamics (SPH) proposed by Müller et al. SPH uses the temperature item to generate particles with different speeds and simulate bloodstream. The multi-thread parallel technology compute unified device architecture (CUDA), which is implemented on GPUs, is used to rapidly solve the control equation of the particle and the blood-solid interactions. Thus, this method allows real-time bleeding based on a uniformly spaced grid. Result The proposed method can simulate bleeding after a cut. It can also simulate blood flowing over complex obstacles in a surgical simulator. The method only takes 20 ms when the particle number is 9 000 and has 20.15 times faster computation than the central processing unit implementation. Thus, real-time bleeding simulation can be obtained with the method using a large particle number. Conclusion The proposed method is flexible and capable of simulating real-time surgical bleeding.