基于模型约简和GPU加速的软组织实时仿真

Real-time Simulation of Body's Soft Tissue Based on Model Reduction and Acceleration of GPU

为了得到虚拟外科手术中人体软组织的实时变形仿真,采用了基于模态分析的模型约简方法,同时结合了GPU加速技术,实现了人体软组织的实时仿真计算,为虚拟手术提供了良好的人机交互。模型约简以有限元理论为基础,将软组织模型的动力学方程投射到约简子空间,减少系统自由度。在实时仿真过程中,首先需在子空间中对低维动力学模型进行数值计算,然后利用GPU通用计算技术来加速重建原始空间中的形变向量。以心脏大变形实时仿真为例,验证方法的性能,实验结果表明心脏的变形仿真可以在很高的刷新率下运行。提出的模型约简和GPU加速方法,为人体软组织的变形实现了实时仿真,为虚拟手术提供了理论基础和技术支撑。

This paper presents a novel method that primarily based on modal analysis and model reduction to implement real - time surgery simulation targeting soft tissues. Model reduction is a popular method for deformable models, mainly because it can project the high - dimensional finite element system into a suitably chosen low - dimensional subspaee, which can substantially reduces the system＇s degrees of freedom. In addition, we also incorporate the power of General - purpose Graphic Processing Unit （ GPGPU ） to reconstruct the deformation vector into the original space, which leaves more rooms on CPU for additional computation such as collision detection, etc. The experi- ment results indicate that the cardiac model presented in this paper can run at high updating rates stably and smoothly.