Closed-Loop Stimulus Optimization to Increase Communication Efficiency in Brain-Computer Interfaces

摘要 辅助设备，如增强和替代通信(AAC)系统，由患有以下疾病的人使用 沟通和运动障碍，如肌萎缩侧索硬化症(ALS，俗称LOU) 格里克病)，与它们的环境交流和互动。有各种商业上的 可用的AAC设备，可通过触摸、切换、头部跟踪和眼睛凝视等访问方法进行控制； 然而，当持续的肌肉控制变得更加困难或不可能时，这些进入方法变得困难或不可能使用 具有挑战性或自愿性的运动控制会丢失。有脑机接口(BCI)通信 感觉神经系统，如P300拼写器，使用感觉刺激来诱导并检测感觉神经 脑电(EEG)数据中的响应，这些通信辅助设备不需要任何马达 在受影响的个人方面进行控制。然而，刺激驱动的准确性和拼写速度 BCI是次优的，因为依赖感觉刺激的固有限制，这是 产生高度可变的神经反应，以及处理固有噪声的脑电数据的必要性 提取控制脑-机接口所需的相关神经信息。当前的BCI通信速率可以 可能会通过利用来自用户的信息的闭环优化技术来改进 对先前刺激的响应，以优化调整BCI系统的参数，以实现预期的目标 在不确定条件下最大化系统性能。可以使用闭环策略来选择 在给定被测量的神经反应的情况下最大限度地提供用户意图信息的刺激， 我假设，这种数据自适应的刺激选择方法将最大限度地减少BCI决策错误和 实现更好的设备控制。传统的脑机接口使用开环刺激控制方法作为刺激 演示计划通常是提前设定的，或者是随机发生的，而且进展有限 BCI中的闭环式刺激范例。我提出的这项研究的目标是调查 一种新的闭环刺激选择算法的可行性，该算法将优化BCI刺激呈现 基于测量的脑电数据和BCI系统对用户意图的信念实时调度， 概念验证在P300 BCI拼写器中演示。《特定目标1》将首先开发和测试这部小说 非残疾人队列中的算法，以利用非残疾人参与者的时间效率和实用性 研究评价闭环刺激选择算法的实时可行性和潜在实用性。 特定目标2将在ALS患者中测试闭环刺激选择算法，以评估 该算法在临床相关队列中的性能。成功地开发和测试了 提出的闭环刺激选择算法在一个具有挑战性的系统中，如P300 BCI拼写器具有 有可能引发向BCI控制和其他应用的闭环方法的范式转变 其中，实时优化系统参数以提高整体系统性能可能是有益的。

Abstract Assistive devices such as augmentative and alternative communication (AAC) systems are used by people with communication and motor disabilities, such as amyotrophic lateral sclerosis (ALS, commonly known as Lou Gehrig’s disease), to communicate and interact with their environment. There are various commercially available AAC devices that are controlled by access methods such as touch, switch, head tracking and eye gaze; however, these access methods become difficult or impossible to use when sustained muscle control is more challenging or voluntary motor control is lost. There are brain-computer interface (BCI) communication systems, such as the P300 speller, that use sensory stimulation to elicit and then detect sensory neural responses in electroencephalography (EEG) data, and these communication aids do not require any motor control on the part of the affected individual. However, the accuracies and spelling speeds of stimulus-driven BCIs are suboptimal due to the inherent limitations associated with relying on sensory stimulation, which generates highly variable neural responses, as well as the necessity of processing inherently noisy EEG data to extract the relevant neural information that is needed to control the BCI. Current BCI communication rates can potentially be improved with closed-loop optimisation techniques that exploit information from the user’s responses to previous stimuli to optimally tune the BCI system’s parameters to achieve the desired goal of maximising system performance under conditions of uncertainty. A closed-loop strategy can be used to select stimuli that are maximally informative of the user’s intent given the neural responses that are being measured, and I hypothesise that this data-adaptive approach to stimulus selection will minimise BCI decision errors and achieve better device control. Conventional BCIs use open-loop stimulus control methods as the stimulus presentation schedule is typically set in advance or occurs randomly, and there has been limited development of closed-loop stimulus paradigms in BCIs. The goal of the research that I propose is to investigate the feasibility of a novel closed-loop stimulus selection algorithm that will optimise the BCI stimulus presentation schedule in real-time based on the measured EEG data and the BCI system’s belief about the user’s intent, with proof-of-concept demonstrated in the P300 BCI speller. Specific Aim 1 will initially develop and test the novel algorithm in a non-disabled cohort to leverage the time efficiency and practicality of non-disabled participant studies to evaluate the real-time feasibility and potential utility of the closed-loop stimulus selection algorithm. Specific Aim 2 will test the closed-loop stimulus selection algorithm in individuals with ALS to assess the performance of the algorithm in a clinically relevant cohort. The successful development and testing of the proposed closed-loop stimulus selection algorithm in a challenging system such as the P300 BCI speller has the potential to instigate a paradigm shift towards closed-loop methods for BCI control and other applications where optimising system parameters in real-time to improve overall system performance could be of benefit.