综合使用CPU和GPU的实时手术仿真系统并行框架  

Parallel Framework Using Both CPU and GPU for Real-Time Surgical Simulation System

作　　者：贾世宇[1] 潘振宽[1] 

机构地区：[1]青岛大学信息工程学院,青岛266071

出　　处：《系统仿真学报》2014年第2期332-338,共7页Journal of System Simulation

基　　金：山东省自然科学基金(ZR2009GM014)

摘　　要：使用多线程技术实现了任务并行化的实时手术仿真系统框架。系统由主控制、力反馈控制、变形计算和GPU控制四大主线程和若干碰撞辅助线程构成。主控制线程处理手术工具与软组织的交互操作、碰撞检测和碰撞响应。力反馈控制线程处理力反馈设备状态读入和反馈力输出。变形计算线程完成基于CPU的软组织变形计算。GPU控制线程处理三维场景图形渲染和基于GPU的软组织变形计算。碰撞辅助线程与主控制线程协同运作完成碰撞检测的并行处理。系统在数据结构和运行机制上进行了特殊设计,使得切割操作可以和GPU加速的变形计算并行运作。对若干测试场景的仿真结果表明,并行框架可以提高系统运行速度,并且运行稳定正常,无死锁现象。Task parallel framework was designed for real-time surgical simulation system using multi-threading technique. The system consists off our main threads： main control haptic device management, deformation calculation and GPU control as well as several auxiliary collision threads. Main control thread performs interaction between surgical tool and soft tissue, collision detection and response. Haptic device management thread performs haptic device state input and force output. Deformation calculation thread performs CPU-based deformation calculation of soft tissue. GPU control thread performs 3D scene rendering and GPU-based deformation calculation of soft tissue. Auxiliary collision threads cooperate with main control thread to perform parallel collision detection. Special designs were made for data structures and running mechanisms in order to make cutting operation and GPU-based deformation calculation run in parallel Results from several simulation test scenes show that the framework can increase system running speed and is stable without dead-locks.

关 键 词：实时手术仿真 软组织变形 软组织切割 GPU加速 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]