Non-invasive EEG-based Continuous Three-dimensional Brain-Computer Interface

总结 仅在美国就有大约500万例瘫痪病例，其中大约140万例是 脊髓损伤（SCI）。在中度到极端的SCI病例中，患者必须严重依赖他人 从喂食到洗澡的护理。虽然这些人强烈渴望增加自主权， 目前存在的治疗选择非常少。近年来，植入侵入性神经脑计算机 脑机接口（BCI）已经显示出通过提供替代的非 大脑的生理交流通道。然而，这些BCI带有大量的短期和长期 术语健康风险，因为手术植入本身就是一种风险，并伴有免疫反应的风险 和感染相比之下，虽然非侵入性脑电图（EEG）脑机接口不会对患者造成重大风险， 个人，他们受到空间分辨率的限制。只有通过新的空间滤波器的最新进展， 独特的，直观的任务，和复杂的训练范例，脑电图脑机接口已扩展到三维 离散任务虚拟光标控制和二维连续虚拟光标和机械臂控制。的 本文提出的研究旨在进一步研究这种新型控制的动态特性和调谐特性 信号，更好地了解在非侵入性BCI期间组合认知任务对认知负荷的影响， 联合收割机将这两种先进技术结合起来，用于实际三维连续虚拟光标控制的训练和演示。 这项工作的主要假设是，通过结合我们对运动想象的理解， 发展我们对外显空间注意力（OSA）作为一种直观控制的神经科学理解 信号，个人将能够鲁棒地，连续地控制三维虚拟光标 利用无创脑电图脑机接口。为了实现这一目标，提出了两个具体目标。首先，我会 研究OSA的空间组织，以及它对使用高密度 脑电图我将使用反向编码模型来识别空间地图，这些模型的灵感和指导来自 侵入性神经科学研究描绘了视觉空间注意力的调谐概况，以预测用户的轨迹 在空间中的注意力，给出了一个主题的参考框架。我将另外通过头部固定来确定， 旋转实验，是否外显空间注意的表征是头部或身体为中心。其次我 将建立一个新的三维连续追踪训练范式，作为评估 三维控制，并将调查多模态控制策略对认知负荷和控制的影响 实力拟议工作的成功完成将对BCI社区产生重大影响， 特别是通过促进非侵入性技术向临床应用的转化， 提高SCI患者的生活质量。

SUMMARY Of the roughly 5 million cases of paralysis in the United States alone, approximately 1.4 million of these are due to spinal cord injury (SCI). In moderate to extreme cases of SCI, patients must rely significantly on others for care ranging from feeding to bathing. While these individuals hold a strong desire for increased autonomy, very few treatment options currently exist. In recent years, the implantation of invasive neural brain computer interfaces (BCI) has shown promise in increasing patient independence by providing an alternative non- physiological communication channel for the brain. However, these BCIs come with substantial short and long term health risks, as the surgical implantation is in itself a risk, and is accompanied by risks of immune responses and infections. In contrast, while noninvasive electroencephalography (EEG) BCIs pose no major risks to the individuals, they are limited by their spatial resolution. Only through recent advancements in novel spatial filters, unique, intuitive tasks, and complex training paradigms have EEG BCIs been extended to three-dimensional discrete task virtual cursor control and two-dimensional continuous virtual cursor and robotic arm control. The research proposed here aims to further investigate the dynamics and tuning properties of the novel control signals, better understand the effects on cognitive load of combined cognitive tasks during noninvasive BCI, and combine the two advances for the training and demonstration of practical 3D continuous virtual cursor control. The main hypothesis of this work is that by combining our understanding of motor imagery and further developing our neuroscientific understanding of overt-spatial attention (OSA) as an intuitive control signal, individuals will be able to robustly, continuously control a virtual cursor in three dimensions utilizing noninvasive EEG BCI. In order to accomplish this, two specific aims are proposed. Firstly, I will investigate the spatial organization of OSA, as well as its dependence on user head orientation using high density EEG. I will identify spatial maps using inverted encoding models, inspired and guided by knowledge gained from invasive neuroscientific studies delineating the tuning profiles of visual spatial attention, to predict a user’s locus of attention in space given a subject’s frame of reference. I will additionally determine via a head fixation and rotation experiment, whether the representation of overt spatial attention is head or body centric. Secondly, I will establish a novel three dimensional continuous pursuit training paradigm as a testing ground for evaluating 3D control, and will investigate the effects multimodal control strategies have on cognitive load and control strength. The successful completion of the proposed work will have a significant effect on the BCI community, particularly by contributing towards the translation of non-invasive technologies towards clinical use and improving the quality of life of SCI patients.