稀疏加权算法与GREIT算法在颅脑电阻抗成像中的对比研究  

Comparative Study on the Application of Sparse Constrained Algorithm and GREIT in Brain Electrical Impedance Tomography

作　　者：李昊庭 徐灿华[1] 刘本源[1] 杨琳[1] 董秀珍[1] 付峰[1] 

机构地区：[1]第四军医大学生物医学工程系,陕西西安710032

出　　处：《中国医疗设备》2016年第11期23-27,共5页China Medical Devices

基　　金：国家自然科学基金(51477176);军队重大课题(AWS14C006)

摘　　要：稀疏加权算法(L1算法)与GREIT算法是近年来两种较为热门的颅脑电阻抗成像(Electrical Impedance Tomography,EIT)图像重建优化算法,基于不同的数学模型,这两种算法性能不同。为了改善颅脑EIT重建图像质量并为其算法优选提供依据,本文基于三维颅脑仿真模型开展了仿真研究,对比了传统二阶范数加权算法(L2算法)、稀疏加权算法以及GREIT算法在颅脑电阻抗图像重建中的性能。仿真结果表明,相比传统二阶范数加权算法,两种优化的算法对颅脑EIT均有改善,而稀疏加权算法在图像噪声、形变误差、位置误差3项指标的评价中性能最优。在阈值函数的作用下,稀疏加权算法经多步迭代抑制了图像噪声,突出了重建目标。稀疏加权算法可以大大改善颅脑EIT效果,适用于颅脑EIT且对未来颅脑电阻抗重建算法扩展研究有重要参考价值。Sparse constrained and GREIT algorithms were two hot optimized reconstruction algorithmsfor brain Electrical Impedance Tomography （EIT）. With different mathematical models, these twoalgorithms had different performances. To give recommendations to brain EIT algorithm selection, basedon the three dimensional head model, a comparison of the performances of conventional quadratic normregularization algorithm （L2 algorithm）, sparse constrained algorithm and GREIT algorithm was madein this paper. On the basis of evaluation results, it could be concluded that sparse constrained algorithmand GREIT demonstrated better performances than L2 algorithm in image noise, location error and shapeerror. With threshold function, the sparse constrained algorithm suppressed the image noise by multi-stepiteration, and highlighted the reconstruction object. Sparse constrained algorithm could greatly improvethe performances of brain EIT and was suitable for brain EIT, which was of great reference for theresearches on the brain EIT reconstruction algorithm EIT in the future.

关 键 词：颅脑电阻抗断层成像 动态成像 图像重建算法 算法比较 三维颅脑模型 

分 类 号：R318[医药卫生—生物医学工程] TM934.7[医药卫生—基础医学]