摘要
运动想象脑-机接口(MI-BCI)可解码用户运动意图,为无法自主运动患者提供一种额外交互控制通道,辅助或改善其生活方式。针对现有下肢MI-BCI分类性能较低等关键问题,引入了体感电刺激(ES)用于下肢MI-BCI构建混合范式(MI+ES),并与传统单一范式(MI)对比。共20名年轻健康右利手受试参与实验,5名参与最优诱发频率验证试验,15名参与正式实验。随后采集了参与正式实验的15名受试不同条件下脑电(EEG)数据,应用傅里叶变换(FFT)和事件相关谱扰动(ERSP)算法提取EEG频域响应、时频特征等,并计算alpha(8~14 Hz)、低beta(15~24 Hz)和高beta(25~35 Hz)等多频段能量变化。此外,分别探索了MI/(MI+ES)条件、共空间模式(CSP)/基于多频率成分的共空间模式(FBCSP)特征提取方法对下肢MI-BCI系统分类性能的影响。结果表明,引入体感电刺激策略可诱发明显的SSSEP特征,MI+ES条件分类准确率较单一MI条件有显著性提升(P<0.001),且应用FBCSP方法的系统分类准确率显著优于经典CSP方法(P<0.01):CSP特征提取方法下MI+ES条件的平均分类准确率为70.2%,其中受试S15的分类准确率达84.2%;FBCSP方法下的平均分类准确率为71.7%,受试S15的分类结果达到90%。初步证实了受试在体感电刺激条件下可诱发出明显的SSSEP特征,而且其融合MI可有效提升下肢MI-BCI分类性能,可支撑下肢MI-BCI系统的实用化进程,也为外周神经相关体感刺激调控方法的优化设计提供了新的技术思路。
Motor imagery(MI)-based brain-computer interface(MI-BCI)can decode motor intention of users,providing an additional interactive manner for patients who are unable to exercise autonomously and improving their lifestyle.To solve the key problem of low classification performance of lower limb MI-BCI,we designed a hybrid paradigm,i.e.MI joint somatosensory electrical stimulation(MI+ES)inducing steady state somatosensory evoked potential(SSSEP)for MI-BCI of lower limb.And the performance of MI+ES was compared with the traditional single paradigm(MI).Twenty right-handed healthy subjects were recruited to participate in the experiment,five of them participated in the verification test of optimal induced frequency and fifteen participated in the formal experiment.EEG data of the fifteen subjects were recorded under different conditions.Fast Fourier transform(FFT)and event-related spectral perturbation(ERSP)were used to extract EEG frequency domain response and time-frequency features.The multi-frequency power changes were calculated at alpha(8~14 Hz),low beta(15~24 Hz)and high beta(25~35 Hz)bands.In addition,the performance of lower limb MI-BCI was explored under different conditions of MI/MI+ES and feature extraction methods of CSP/FBCSP.Results showed that the somatosensory electrical stimulation strategy could induce obvious SSSEP features.The classification accuracy of MI+ES condition was significantly improved in reference to the single MI condition(P<0.001).The classification performance based on FBCSP method was significantly better than that of classical CSP method(P<0.01),the classify accuracy of CSP was 70.2%under MI+ES condition,while the accuracy of subject S15 was 84.2%.And the accuracy of FBCSP was71.7%,the accuracy of subject S15 was 90%.In conclusion,this study preliminarily confirmed that the SSSEP could be evoked by the somatosensory electrical,and the hybrid paradigm could effectively improve the classification performance of lower limb MI-BCI,which could promote the practical development,even provide
作者
张力新
常美榕
王仲朋
陈龙
明东
Zhang Lixin;Chang Meirong;Wang Zhongpeng;Chen Long;Ming Dong(Academy of Medical Engineering and Translational Medicine,Tianjin University,Tianjin 300072,China;College of Precision Instrumental and Optoelectronic Engineering,Tianjin University,Tianjin 300072,China;Chinese Society of Biomedical Engineering)
出处
《中国生物医学工程学报》
CAS
CSCD
北大核心
2021年第4期429-437,共9页
Chinese Journal of Biomedical Engineering
基金
国家自然科学基金重点项目(81630051)
天津市科技计划项目(17ZXRGGX00020,17ZXRGGX00010)。
关键词
下肢运动想象
脑-机接口(BCI)
稳态体感诱发电位(SSSEP)
事件相关谱扰动
分类识别
lower limb motor imagery
brain-computer interface(BCI)
steady state somatosensory evoked potential(SSSEP)
event-related spectral perturbation
classification and recognition
