ECoG Correlates of Visuomotor Transformation and Application to a Force-Based BCI

描述（由申请人提供）：本提案的长期目标是开发一种新的脑机接口（BCI）算法，用于麻痹肢体的功能性神经肌肉刺激等应用。为了实现这一总体目标，本提案的第一个目的是寻求识别运动皮质区域的特性，这些特性可能有助于我们的脑机接口追求。具体而言，我们的第一个目标将寻求识别当任务末端执行器的运动（即用于完成任务的计算机光标）与导致它的物理运动分离时，运动参数在运动相关皮层区域的运动参数表征中的差异和灵活性。先前的研究表明，运动皮层的高伽马波段ECoG和LFP活动（bbb60 Hz）与手的速度密切相关，而单单元研究表明，运动前皮层区域可能更多地参与到手的运动的计划和/或感知中，而不是物理运动本身。然而，这些研究通常采用线性转换的物理手臂运动的视觉反馈的末端执行器的任务。为了进一步研究皮层区域运动表征的差异，我们将训练三只雄性恒河猴（Macaca mulatta）进行二维中心向外操纵杆任务，其中操纵杆的位置（用作手位置的代理）被重新映射到目标导向的计算机光标的位置、速度或加速度，从而消除操纵杆和光标之间的任何线性相关性，以实现速度和加速度重新映射方案。我们将首先检查在M1、PMd和顶叶区5上植入的阵列的硬膜外ECoG频谱，以寻找与操纵杆或光标的运动参数的相关性，并比较不同重映射方案之间的相关性，以确定特定区域是否与运动的计划/感知方面或物理执行方面更相关。我们的第二个目标是利用这些ECoG表征的特性，并确定猴子是否可以学习在脑机接口（BCI）任务中调制基于力的控制信号。迄今为止，大多数脑机接口实验都利用了基于运动学的控制信号（即大脑信号被转换为物体的位置或速度）。然而，这些控制信号可能不适用于瘫痪肢体的功能性神经肌肉刺激（FNS）等应用，因为个人手臂的准确模型未知或不完整。为了解决这个问题并评估可能更适合FNS的基于动态的控制算法的可行性，我们将在基于力的BCI任务上训练猴子。在这些实验中，高伽马（75-105赫兹）的ECoG活动将被映射到计算机光标上的虚拟力上，在一个类似中心向外的任务中，计算机光标将被用来“抓取”和移动有质量的外围目标。虚拟环境还将结合简单的现实世界物理，如重力和变质量，以评估一个对象如何能够很好地适应现实生活中的这些扰动，并使用感兴趣的控制算法。

DESCRIPTION (provided by applicant): The long-term objective of this proposal is to develop a novel brain computer interface (BCI) algorithm targeted for use in applications such as functional neuromuscular stimulation of paralyzed limbs. Toward this overall goal, the first aim of this proposal seeks to identify properties of motor cortical areas that may aid in our BCI pursuits. Specifically, our first objective will seek to identify differences and flexibility in electrocorticography (ECoG) representations of movement parameters across motor-associated cortical areas when movement of the end-effector for a task (i.e. a computer cursor used to complete the task) is dissociated from the physical movement leading up to it. Previous studies have shown that high gamma band ECoG and LFP activity (>60 Hz) over motor cortex is well correlated to hand speed during reaching tasks, while single unit studies have suggested that premotor cortical areas may be more involved in the planning and/or perception of reaching movements than the physical movement itself. However, these studies usually employed a linear transformation of the physical arm movements to visual feedback of the end-effector for the task. To further investigate the differences in representation of movement across cortical areas, we will train three male rhesus monkeys (Macaca mulatta) on a 2-D center-out joystick task in which position of the joystick (used as a proxy for hand position) is remapped to the position, velocity, or acceleration of a goal-oriented computer cursor, thus removing any linear correlations between joystick and cursor for velocity and acceleration remapping schemes. We will first examine the epidural ECoG spectra of arrays implanted over M1, PMd, and parietal area 5 for correlations with movement parameters of the joystick or cursor, and compare these correlations across remapping schemes to identify whether a particular area is more correlated with planning/perceptual aspects of a movement or with its physical execution. Our second objective is to utilize properties of these ECoG representations and determine whether monkeys can learn to modulate a force-based control signal in a brain- computer interface (BCI) task. The majority of BCI experiments to date have utilized kinematic-based control signals (i.e. a brain signal is translated to the position or velocity of an object). However, these control signals may not be suitable for applications such as functional neuromuscular stimulation (FNS) of a paralyzed limb where an accurate model of an individual's arm is unknown or incomplete. To address this problem and assess the feasibility of a dynamics-based control algorithm that may be more appropriate for FNS, we will train monkeys on a force-based BCI task. In these experiments, high gamma (75-105 Hz) ECoG activity will be mapped to a virtual force on a computer cursor in a center-out like task in which the computer cursor will be used to "grab" and move peripheral targets with mass. The virtual environment will also incorporate simple real-world physics such as gravity and variable masses to assess how well a subject might be able to adapt to these perturbations in real-life with the control algorithm of interest.