Neural Encoding and Connectivity in human speech perception

描述（由申请人提供）：定义支持言语和语言功能的基本神经机制是临床研究失语或阅读障碍等沟通障碍的基本挑战。更好地理解语言的神经基础也可能导致神经假肢装置的发展，以使影响沟通能力的神经系统疾病（例如，ALS或中风）丧失功能。要实现这些目标，必须解决两个关键挑战。首先，大脑皮层的活动是如何表现语音的？尽管说话者和环境条件存在很大差异，但人类仍能流畅地理解语音，但支持这种不变识别能力的潜在神经表征从根本上是未知的。第二，大脑皮层言语表征的解剖学基础是什么？理论模型和非人类动物数据表明，听觉对象识别可能是分层组织的，但这种结构是否存在于人类大脑中尚不清楚。该项目将通过研究神经外科患者的微电极阵列对语音的颅内（皮质电图，ECoG）反应，专注于神经表征和连通性这两个关键问题。在指导的K99阶段，第一个具体目标是使用神经编码模型方法研究语音在高阶听觉皮层中的神经表征。神经编码模型定量地描述了特定大脑区域编码的语音特征，并预测了对新语音刺激的神经反应。将通过比较语音模型和基于语音谱图的线性光谱时间模型来测试是否存在绝对语音语音表示。语音模型代表语音级别类别之间的声学不变性。在独立的R00阶段，第二个具体目标将使用皮质电刺激（ECS）和功能连接分析来研究支持语音感知的神经回路的连通性。ECS直接绘制受刺激脑部位的解剖连通性，而功能连通性分析则确定这些连接是如何发生的

DESCRIPTION (provided by applicant): Defining the basic neural mechanisms supporting speech and language function is a fundamental challenge for clinical insights into communication disorders such as aphasia or dyslexia. Better understanding of the neural basis of speech may also lead to development of neural prosthetic devices for disabling neurological disorders affecting communication ability (e.g., ALS or stroke). To achieve these goals, two key challenges must be addressed. First, how is speech sounds represented by cortical activity? Humans fluidly understand speech despite large variations in speakers and environmental conditions, but the underlying neural representations that support this invariant recognition ability are fundamentally unknown. Second, what is the anatomical substrate of cortical speech representation? Theoretical models and nonhuman animal data suggest auditory object recognition may be organized hierarchically, but it remains unknown if this architecture is present in the human brain. This project will focus on these two key questions, neural representation and connectivity, by investigating intracranial (electrocorticographic, ECoG) responses to speech measured with microelectrode arrays in neurosurgical patients. During the mentored K99 phase, the first specific aim will investigate the neural representation of speech in higher order auditory cortex using a neural encoding model approach. A neural encoding model describes quantitatively what speech features are encoded by specific brain areas and predicts the neural response to novel speech stimuli. The presence of a categorical phonetic speech representation will be tested by comparing a phonetic model, which represents acoustic invariance among phone-level categories, to a linear spectrotemporal model based on the speech spectrogram. During the independent R00 phase, the second specific aim will investigate the connectivity of neural circuits supporting speech perception using electrical cortical stimulation (ECS) and functional connectivity analysis. ECS directly maps anatomical connectivity of stimulated brain sites, while functional connectivity analysis identifies how these connections are modulated during speech perception. Comparison of the connectivity maps to the feature selectivity of fitted encoding models (Aim 1) will provide a comprehensive view of the functional organization of speech representation in higher order auditory cortex. Understanding the representation and connectivity of speech recognition in human cortex has significant implications for a number of health applications. Accurate encoding models can be used to decode speech from neural activity and form the basis of prosthetic devices for communication. Furthermore, mapping the functional organization of speech will allow more precise determination of critical speech sites during neurosurgical procedures and will provide insights into key brain areas and circuits underlying communication disorders such as aphasia and developmental dyslexia.