基于fNIRS技术的脑机接口研究

Investigation of Brain-Computer Interface Using Functional Near-Infrared Spectroscopy

功能性近红外光谱技术(functional near-infrared spectroscopy,fNIRS)作为一种无损光学脑成像技术,可用于构建脑机接口(brain-computer interface,BCI)以识别操作者肢体控制意图.利用fNIRS技术测量11位参试者的手臂伸展、腿部伸展和手指敲击的前额皮层(prefrontal cortex,PFC)和运动功能皮层(motor cortex,MC)的血氧变化信号,并利用fNIRS的生理特征和支持向量机建立fNIRS-BCI分类器.结果表明手臂伸展、手指敲击和腿部伸展的四分类fNIRS-BCI平均正确率分别为89.32%,、88.66%,和91.35%,;fNIRS-BCI的运动想象动作的分类正确率不低于运动执行动作;3种任务范式的混淆矩阵分析结果表明:运动想象诱发的脑功能活动与同侧肢体的运动执行、对侧肢体的运动想象活动产生混淆,3种任务范式的同侧运动想象和运动执行的血氧数据检验结果存在显著差异.因此,fNIRS-BCI能有效识别运动想象和运动执行活动,且运动想象和运动执行活动的血氧数据变化具有可分性.

Functional near-infrared spectroscopy(fNIRS)is a non-invasive optical brain-imaging technology whichcould be applied in brain-computer interface(BCI)to recognize motor intention in brain.The study used fNIRS tomeasure hemodynamic variation in prefrontal cortex(PFC)and motor cortex(MC)while11participants were completingthree different movement tasks,that is,hand movement,leg movement and finger tapping.In order to discriminatebetween motor execution(ME)and motor imagery(MI),three fNIRS-BCI classifiers were established usingfNIRS feature and support vector machine.The result shows that three four-class fNIRS-BCI classifiers,corresponding to hand movement,finger tapping and leg movement,achieve average classification accuracy of89.32%,,88.66%,and91.35%,respectively.The classification accuracy of MI by fNIRS-BCI is not lower than that ofME.The analysis of confusion matrix for three experiment paradigms shows that cerebration induced by MI confuseswith ME of ipsilateral limb and MI of offside limb.The concentration of hemodynamic changes of MI and ME for thesame limb in the three paradigms shows significant difference.Hence,the present study verifies the feasibility offNIRS-BCI to differentiate between MI and ME in brain,and hemodynamic changes in response to MI and ME areseparable.