Clinic Interactions of a Brain-Computer Interface for Communication

 描述（由申请人提供）：对于那些无法依靠言语或书写来表达自己的严重言语和身体障碍（SSPI）的人来说，脑机接口（BCI）的交流前景正在成为现实。虽然大多数研究工作致力于技术开发以解决稳定性、可靠性和/或分类问题，但随着 SSPI 患者及其家庭/护理团队在新手或长期试验期间评估系统，临床和行为挑战变得越来越明显。 RSVP Keyboard(tm) BCI 转化研究团队的目标是通过创新的工程设计来解决功能性 BCI 使用过程中提出的临床挑战，从而增强这种新型辅助技术的潜力。提出了四个具体目标：（1）开发BCI通信应用套件（BCI-CAS），为SSPI患者提供一组能够满足其语言/读写能力的语言模块； (2) 开发改进的统计信号模型，用于个性化特征提取、伪影/干扰处理，以及从生理信号中提取稳健、准确的意图证据； (3) 开发改进的语言模型和刺激序列优化方法； (4) 评估影响 BCI-CAS 学习和表现的认知变量。提出了五个语言模块，它们依赖于多模态证据融合框架来进行基于模型的上下文感知最佳意图推理：RSVP Keyboard(tm) 生成拼写；回复短信； RSVP 上下文打字； RSVP 上下文图标输入；以及 SSVEP 的二元是/否响应。当前 RSVP Keyboard(tm) 和 SSVEP 系统的可用性数据推动了所有提议的目标。用户选择一个语言模块，BCI 系统根据用户适应、意图推断和个性化语言建模来优化每个人的性能。独特的模拟功能推动系统参数的个性化。 BCI 定制工作的稳健性由具有 SSPI 和神经典型对照的成年人以迭代方式不断评估。解决注意力认知结构的三个干预计划（特定过程的注意力训练、正念冥想训练和新颖的刺激呈现）的效果将通过假设驱动的单主题设计来实现。 30 名年龄在 21 岁及以上、患有 SSPI 的参与者将被纳入家庭干预措施中。通过测量信息传输率 (ITR)、用户满意度和内在用户因素，我们将确定影响患有神经退行性或神经发育疾病的成年人的 BCI 基台获取和表现的学习策略。翻译团队包括（1）信号处理（Erdogmus）； （2）临床神经生理学（Oken）； （3）自然语言处理（Bedrick/Gorman）； (4) 辅助技术（Fried-Oken）。我们继续依赖坚实的贝叶斯基础和理论框架：ICF 残疾分类（WHO，2001）、AAC 参与模型（Beukelman 和 Mirenda，2013）以及人与技术匹配模型（Scherer，2002）。

 DESCRIPTION (provided by applicant): The promise of brain-computer interfaces (BCI) for communication is becoming a reality for individuals with severe speech and physical impairments (SSPI) who cannot rely on speech or writing to express themselves. While the majority of research efforts are devoted to technology development to address problems of stability, reliability and/or classification, clinical and behavioral challenges are becoming more apparent as individuals with SSPI and their family/care teams assess the systems during novice or long-term trials. The objective of the RSVP Keyboard(tm) BCI translational research team is to address the clinical challenges raised during functional BCI use with innovative engineering design, thereby enhancing the potential of this novel assistive technology. Four specific aims are proposed: (1) to develop a BCI Communication Application Suite (BCI-CAS) that offers a set of language modules to people with SSPI that can meet their language/literacy skills; (2) to develop improved statistical signal models for personalized feature extraction, artifact/interference handling, and robust, accurate intent evidence extraction from physiologic signals; (3) to develop improved language models and stimulus sequence optimization methods; and (4) to evaluate cognitive variables that affect learning and performance of the BCI-CAS. Five language modules are proposed that rely on a multimodal evidence fusion framework for model-based context-aware optimal intent inference: RSVP Keyboard(tm) generative spelling; RSVP texting; RSVP in-context typing; RSVP in-context icon typing; and binary yes/no responses with SSVEPs. Usability data on the current RSVP Keyboard(tm) and SSVEP system drive all proposed aims. Users select a language module, and the BCI system optimizes performance for each individual based on user adaptation, intent inference, and personalized language modeling. A unique simulation function drives individualization of system parameters. The robustness of the BCI customization efforts are evaluated continually by adults with SSPI and neurotypical controls in an iterative fashion. The effect of three intervention programs that address the cognitive construct of attention (process-specific attention training, mindfulness meditation training and novel stimulus presentations) will be implemented through hypothesis-driven single subject designs. Thirty participants, ages 21 years and older with SSPI will be included in home-based interventions. By measuring information transfer rate (ITR), user satisfaction, and intrinsic user factors, we will identify learning strategies that influence BCI sill acquisition and performance for adults with neurodegenerative or neurodevelopmental conditions. The translational teams include (1) signal processing (Erdogmus); (2) clinical neurophysiology (Oken); (3) natural language processing (Bedrick/Gorman); and (4) assistive technology (Fried-Oken). We continue to rely on a solid Bayesian foundation and theoretical frameworks: ICF disability classification (WHO, 2001), the AAC model of participation (Beukelman & Mirenda, 2013) and the Matching Person to Technology Model (Scherer, 2002).