Brain communication in the locked-in state
Brain communication in the locked-in state
复制标题
DOI：
10.1093/brain/awt102
复制
发表时间：
2013-06-01
期刊：
影响因子：
14.5
通讯作者：
Birbaumer, Niels
中科院分区：
文献类型：
--
作者：
De Massari, Daniele;Ruf, Carolin A.;Birbaumer, Niels
处于完全锁定状态的患者没有任何交流手段，他们代表了过去15年来脑机接口研究的目标人群。尽管已经测试了不同的范例并且使用了不同的生理信号，但是迄今为止没有充分记录的完全锁定状态患者能够在延长的时间段内控制脑机接口。我们引入巴甫洛夫语义条件反射，使基本的通信完全锁定的状态。这种新的范例是基于语义条件反射的神经电或任何其他生理信号的在线分类，以区分隐蔽（认知）的“是”和“否”的反应。该范例包括作为条件刺激的肯定和否定陈述的呈现，而非条件刺激包括皮肤的电刺激与肯定陈述配对。三名患有晚期肌萎缩侧索硬化症的患者参与了一段时间，其中一名处于完全锁定状态，另外两名处于锁定状态。通过听觉oddball程序评估患者的警觉水平，以研究警觉水平与分类器性能之间的相关性。所有患者脑电信号慢皮层成分的平均在线分类准确率均在机会水平附近。在离线分类程序中使用非线性分类器导致在一个锁定状态患者中实现70%正确分类的准确性的实质性提高。尽管认知处理能力完好无损，但在37个会话中，完全锁定状态患者的可靠性能水平并没有均匀地实现，但在某些会话中，通信准确性达到了70%。提出了范式修正。警觉性的快速下降表明注意力的变化或昼夜节律的变化是脑-机接口通信的锁定状态和完全锁定状态的重要因素。
Patients in the completely locked-in state have no means of communication and they represent the target population for brain-computer interface research in the last 15 years. Although different paradigms have been tested and different physiological signals used, to date no sufficiently documented completely locked-in state patient was able to control a brain-computer interface over an extended time period. We introduce Pavlovian semantic conditioning to enable basic communication in completely locked-in state. This novel paradigm is based on semantic conditioning for online classification of neuroelectric or any other physiological signals to discriminate between covert (cognitive) 'yes' and 'no' responses. The paradigm comprised the presentation of affirmative and negative statements used as conditioned stimuli, while the unconditioned stimulus consisted of electrical stimulation of the skin paired with affirmative statements. Three patients with advanced amyotrophic lateral sclerosis participated over an extended time period, one of which was in a completely locked-in state, the other two in the locked-in state. The patients' level of vigilance was assessed through auditory oddball procedures to study the correlation between vigilance level and the classifier's performance. The average online classification accuracies of slow cortical components of electroencephalographic signals were around chance level for all the patients. The use of a non-linear classifier in the offline classification procedure resulted in a substantial improvement of the accuracy in one locked-in state patient achieving 70% correct classification. A reliable level of performance in the completely locked-in state patient was not achieved uniformly throughout the 37 sessions despite intact cognitive processing capacity, but in some sessions communication accuracies up to 70% were achieved. Paradigm modifications are proposed. Rapid drop of vigilance was detected suggesting attentional variations or variations of circadian period as important factors in brain-computer interface communication with locked-in state and completely locked-in state.