【作者】 辛红；

【导师】 屈晓超；

【作者基本信息】 西安电子科技大学 ， 生物医学工程， 2014， 硕士

【摘要】 激光技术的出现使得光学成像技术在现代医学和生物学中的应用越来越广泛。随着光学成像技术的不断发展,迫切需要更为准确地认识光子在复杂生物组织中的传输过程。蒙特卡洛方法(Monte Carlo Method,MCM)能够精确地描述光在生物组织中的传输过程或规律,被研究人员誉为求解光传输问题的金标准。然而,由于随机统计特性,MCM执行效率偏低,阻碍了MC方法在医学领域更深层次的推广。本文以提高MCM仿真效率为目标,建立了一个基于网络传输的光学分子影像仿真平台并提出相应的仿真加速方法。本文针对现有光学分子影像仿真平台使用复杂并且效率低下的问题,建立了基于网络传输的光学分子影像仿真平台(MOSE Online Version)。该平台分为前端动态网页、后端光传输蒙卡仿真服务器和显示插件三部分。首先,对平台的整体结构、仿真算法、页面显示及结果展示等部分进行独特设计与实现。其次,为了确保仿真平台的稳定性,设计并完成了一系列的测试工作,包括操作界面测试、功能测试和兼容性测试等。并基于两组相同的物理模型,将基于网络传输的仿真平台得到的仿真结果与本地仿真平台所得到的结果进行对比,验证基于网络传输的光学分子影像仿真平台的准确性。最后,分别运行基于网络传输与本地的仿真平台,进行用户使用效率的对比,结果表明基于网络传输的仿真平台可以显著改善用户使用效率,减轻用户端的使用负担。在多用户多任务访问MOSE Online Version平台服务器时,会造成服务器数据拥堵甚至服务器崩溃的情况。本文提出了基于Hadoop的光传输蒙特卡洛仿真架构来解决这一问题。首先,使用Hadoop架构执行与光传输蒙卡仿真机制相似的PI运算,对该框架执行蒙特卡洛方法的可行性进行验证。其次,建立了基于云计算系统的光传输蒙特卡洛仿真模型。并在该模型基础上,利用多台计算机组成的基于Hadoop架构的MapReduce计算集群,结合光在介质内传输时按步长吸收权值的模型特性,建立了标准化的蒙特卡洛仿真程序接口。最后,将该仿真程序包导入MapReduce计算集群进行蒙特卡洛仿真。这种架构能够很好的调度服务器集群间的仿真任务,从而使各个服务器达到负载均衡,有效解决仿真平台在执行大量计算任务时光传输蒙特卡洛仿真服务器的执行效率及可靠性问题。更多还原

【Abstract】 Laser technology makes the application of optical imaging technology wider and wider in modern medicine and biology. With the development of optical imaging technology, it is urgent to accurately understand the transmission process of light within complex biological tissues. Monte Carlo method(MCM) can accurately describe the optical transmission procedure or law in biological tissues, so it is hailed as the golden standard for solving optical transmission problem by researchers. However, the efficiency of MCM is relatively low due to the randomly statistical properties, which definitely hinders the development of MCM in the medical field. Aiming at improving the low simulation efficiency of MCM, an optical molecular imaging simulation platform based on network transport and and its acceleration method is implemented in this thesis.An optical molecular imaging simulation platform based on network transmission was developed to promote the efficiency and to reduce the complexity of the existing optical molecular imaging simulation platform.The platform included sub-front Dynamic Web-Pages, the back-end Monte Carlo simulation server of light transmission and display plug-in. Firstly, the overall platform structure, simulation algorithm, page display and result show were specifically designed and implemented. Secondly, to guarantee the robustness of the simulation platform based on network transmission, a series of test tasks including operation interface, function and compatibility were designed and conducted. Based on two identical physical models, the simulation data obtained by the simulation platform using network transmission was compared with that by the local simulation platform, thus the accuracy of MOSE based on network transmission was validated. Finally,with a comparison between MOSE and its Online Version, the results gave evidence that MOSE Online Version significantly improved the user efficiency, and reduced clients’ burdens.When the platform server of MOSE Online Version is accessed by multi-user and multitask, it may come up with the data congestion and server crash. To handle this problem, this thesis proposes simulation framework of optical transmission with Monte Carlo based on Hadoop. Firstly, the PI computation that is similar to optical transmission Monte Carlo simulation theory is conducted to demonstrate its feasibility in Monte Carlo simulation by Hadoop framework. Secondly, the optical transmission Monte Carlo simulation model is built based on cloud computing. Based on the simulation model, combining the multiple Map Reduce computing cluster based on Hadoop framework with the model feature where light is absorbed according to step length as transmitting in medium, the standard Monte Carlo simulation interface is built. Finally, this program package is led into Map Reduce computing cluster to conducted Monte Carlo simulation. This framework can schedule the simulation task among server clusters, to balance the loading, which improves the efficiency of massive calculation of optical transmission Monte Carlo simulation within server and guarantees the reliability.更多还原