基于TensorFlow的并行化FBP方法  

Parallelized FBP Method Based on TensorFlow

作　　者：朱炯滔 王成 ZHU Jiongtao;WANG Cheng(School of Mechanical and Electronic Engineering,Wuhan University of Technology,Wuhan 430070,China)

机构地区：[1]武汉理工大学机电工程学院,湖北武汉430070

出　　处：《数字制造科学》2020年第3期210-213,共4页

摘　　要：针对在卷积神经网络中使用全连层来完成CT图像重建而带来的计算资源占用高、网络参数过多等问题,提出了一种使用TensorFlow提供的自定义数学操作来导入CT重建算法的方法。重建算法为滤波反投影算法(filter back projection,FBP),使用GPU(graphics processing unit)对其进行并行化计算。实验结果表明,该方法可以显著减少网络参数和占用的计算资源,加快网络运算速度。Aiming at solving the problems of high computational consumption and excessive network parameters in convolutional neural networks when using full-layer to complete CT image reconstruction,an imported CT reconstruction algorithms is proposed in this paper.It is designed based on a customised mathematical operation(OP)provided by TensorFlow.The reconstruction algorithm is a filtered back-projection algorithm(FBP),and run as well as optimized on GPU.Experimental results show that this method can significantly reduce the number of network parameters and decrease computational consumption,therefore speeding up the network computing.

关 键 词：TensorFlow 并行化计算 神经网络 CT重建 

分 类 号：R-05[医药卫生] TP391.41[自动化与计算机技术—计算机应用技术]