Saga Norihiko, Doi Atsushi, Oda Teruo, Kudoh Suguru N
Department of Human System Interaction, School of Science and Technology, Kwansei Gakuin University, Sanda, Japan.
Front Neurorobot. 2021 Jan 25;14:607706. doi: 10.3389/fnbot.2020.607706. eCollection 2020.
Non-invasive brain-computer interfaces (BCIs) based on common electroencephalography (EEG) are limited to specific instrumentation sites and frequency bands. These BCI induce certain targeted electroencephalographic features of cognitive tasks, identify them, and determine BCI's performance, and use machine-learning to extract these electroencephalographic features, which makes them enormously time-consuming. In addition, there is a problem in which the neurorehabilitation using BCI cannot receive ambulatory and immediate rehabilitation training. Therefore, we proposed an exploratory BCI that did not limit the targeted electroencephalographic features. This system did not determine the electroencephalographic features in advance, determined the frequency bands and measurement sites appropriate for detecting electroencephalographic features based on their target movements, measured the electroencephalogram, created each rule (template) with only large "High" or small "Low" electroencephalograms for arbitrarily determined thresholds (classification of cognitive tasks in the imaginary state of moving the feet by the size of the area constituted by the power spectrum of the EEG in each frequency band), and successfully detected the movement intention by detecting the electroencephalogram consistent with the rules during motor tasks using a fuzzy inference-based template matching method (FTM). However, the electroencephalographic features acquired by this BCI are not known, and their usefulness for patients with actual cerebral infarction is not known. Therefore, this study clarifies the electroencephalographic features captured by the heuristic BCI, as well as clarifies the effectiveness and challenges of this system by its application to patients with cerebral infarction.
基于常规脑电图（EEG）的非侵入性脑机接口（BCI）局限于特定的仪器放置部位和频段。这些BCI诱发认知任务的某些目标脑电图特征，识别它们并确定BCI的性能，且使用机器学习来提取这些脑电图特征，这使得它们极为耗时。此外，存在一个问题，即使用BCI的神经康复无法接受动态和即时的康复训练。因此，我们提出了一种不限制目标脑电图特征的探索性BCI。该系统不预先确定脑电图特征，而是根据目标运动确定适合检测脑电图特征的频段和测量部位，测量脑电图，针对任意确定的阈值（通过每个频段脑电图功率谱构成的面积大小对想象中脚部运动状态下的认知任务进行分类）仅用大的“高”或小的“低”脑电图创建每个规则（模板），并使用基于模糊推理的模板匹配方法（FTM）在运动任务期间通过检测与规则一致的脑电图成功检测到运动意图。然而，这种BCI获取的脑电图特征尚不清楚，其对实际脑梗死患者的有用性也未知。因此，本研究阐明了启发式BCI捕获的脑电图特征，并通过将该系统应用于脑梗死患者来阐明其有效性和挑战。
