基于ICA方法去除人工耳蜗ERP信号伪迹的研究  

Removal of cochlear implant artifacts of ERP based on independent component analysis

作　　者：闫立丽[1] 张旭[1] 陈雪清[2,3] 傅新星[2,3] 刘斌[1] 钱柏霖 

机构地区：[1]首都医科大学生物医学工程学院临床生物力学应用基础研究北京市重点实验室,北京100069 [2]首都医科大学附属北京同仁医院耳鼻喉科 [3]北京市耳鼻咽喉科研究所北京市耳鼻咽喉头颈科学教育部重点实验室,北京100005

出　　处：《北京生物医学工程》2015年第2期111-117,150,共8页Beijing Biomedical Engineering

基　　金：北京市自然科学基金(3122011)资助

摘　　要：目的人工耳蜗植入者的听觉诱发电位含有较大的伪迹信号,影响了其在人工耳蜗植入后的效果评估。功能成像方法由于安全问题和介入性等特点,不适用于人工耳蜗植入者。本文利用独立成分分析(independent component analysis,ICA)去除人工耳蜗伪迹,为进一步利用听觉诱发电位信号客观评价人工耳蜗植入者言语识别能力和人工耳蜗植入效果提供便利。方法采用经典Oddball模式,分别以言语声/ba/和/da/为标准刺激和偏差刺激,测量人工耳蜗植入者的听觉事件相关电位(eventrelated potential,ERP),采用ICA方法去除ERP信号中人工耳蜗造成的伪迹,并分析其独立成分的时域波形和脑地形图特征。本文对10例人工耳蜗植入6个月的受试者进行ERP测试,并比较了Infomax和Jade两种算法去除人工耳蜗伪迹的效果。结果根据独立成分的时域波形和脑地形图特征,可以将人工耳蜗伪迹对应的独立成分识别出来。人工耳蜗伪迹独立成分的时域波形类似于一个基座,其脑地形图显示在植入侧有较高的电位。去除人工耳蜗伪迹后的ERP波形显示出原始的形态。Infomax算法能够更有效地去除ERP信号中的人工耳蜗伪迹。结论 ICA方法可以有效地将人工耳蜗伪迹从人工耳蜗植入者的ERP信号中分离出来。Objective The auditory evoked potentials（ AEP）of cochlear implant（ CI）users contalned CI-related artifact,which restrict the assessment for implantation effect. Functional imaging methods are not suitable for CI users due to the safety and invasiveness. The primary objective of this study is to investigate noise-reduction of event-related potentials （ ERP ） for CI users based on independent component analysis （ ICA）. It can provide more information on assessment for speech perception and implantation effect of CI users with AEP. Methods The standard（/ba/） and deviant stimuli（/da/） were presented in an oddball paradigm and the ERPs of cochlear implant users were recorded. ICA was used to remove the cochlear implant artifacts in ERP. The characteristic of waveform and braln topographic map of cochlear implant artifacts were＆amp;nbsp;analyzed. Ten CI users with device‘switch on’for six months took part in this experiment and the effect of artifacts removal with Infomax and Jade algorithms were compared. Results The cochlear implant artifacts could be recognized by the characteristic of the waveform and braln topographic maps of independent components. The waveform of cochlear implant artifacts resembled a pedestal and its braln topographic map showed higher potential on the implantation side. The ERP waveforms displayed reasonable morphologies after the removal of cochlear implant artifacts. The Imfomax algorithm was better to remove the CI-related artifacts in ERP. Conclusions The cochlear implant artifacts could be removed from ERP of cochlear implant users based on ICA.

关 键 词：独立成分分析 人工耳蜗 事件相关电位 伪迹 去噪 

分 类 号：R318.04[医药卫生—生物医学工程]