Sensory mechanisms of manual dexterity and their application to neuroprosthetics

项目摘要 手动行为需要来自手的感觉信号，包括触觉和本体感受，如由 由躯体感觉神经传入障碍引起的严重缺陷。手的感觉成分的三个方面 感觉功能知之甚少。首先，触觉的神经基础几乎完全被研究， 被动地传递到皮肤的刺激，排除了对触觉信号如何被调制的任何理解， 与运动指令互动第二，本体感受信号不仅携带关于时间变化的信息， 手的构造，也是关于手动施加的力，但力的本体感受表征 我们对此知之甚少。第三，立体感-物体的三维形状的感觉， 感觉信号来自手-意味着触觉和本体感受信号的整合， 对此知之甚少。主动触觉、手本体感觉和立体感的研究一直受到阻碍 技术障碍。事实上，描述自我产生的与物体的接触是困难的， 这是不可能的，因为有足够的精度跟踪手的运动。为了克服这些障碍，我的团队 已经开发出一种设备，使我们能够测量接触事件-与传感器片覆盖的对象的 表面和跟踪时变的手部姿势-使用基于深度学习的计算机视觉- 动物与物体互动时前所未有的精确度。然后我们描述每个阶段的反应 沿着躯体感觉神经轴，从外周神经穿过皮层。这个新的实验装置将 使我们能够研究躯体感觉的神经基础-特别是当它与手的灵活性有关时- 生态条件良好。 在相关的调查中，我们利用我们对感觉处理的了解来恢复触觉， 仿生手简而言之，我们开发了将仿生手上传感器的输出转换为 电刺激外周神经（对于截肢者）或躯体感觉皮层（对于患有 四肢瘫痪）以唤起有意义的触觉感知。我是仿生方法的主要设计师之一 到人工触摸，它假定模拟自然神经信号的编码算法将产生更多的 直观的触觉感知，从而赋予仿生手更大的灵活性。我们在人工触摸方面的工作 包括三个部分：电刺激的感知相关性的评估， 感官编码算法，以及评估人工触摸对手动行为的好处。的相互作用 基础科学成果和神经工程的努力将为大脑带来更自然的人工触摸- 控制仿生手

PROJECT SUMMARY Manual behavior requires sensory signals from the hand, both tactile and proprioceptive, as evidenced by the severe deficits that result from somatosensory deafferentation. Three aspects of the sensory component of hand sensory function are poorly understood. First, the neural basis of touch has been studied almost exclusively with stimuli delivered passively to the skin, precluding any understanding of how tactile signals are modulated by and interact with motor commands. Second, proprioceptive signals carry information not only about the time-varying conformation of the hand, but also about manually applied forces, but proprioceptive representations of force are poorly understood. Third, stereognosis – the sense of the three-dimensional shape of objects acquired from sensory signals arising from the hand – implies the integration of tactile and proprioceptive signals, a process about which little is known. The study of active touch, hand proprioception, and stereognosis has been hindered by technical obstacles. Indeed, characterizing self-generated contact with objects has been difficult or impossible, as has tracking hand movements with sufficient precision. To overcome these obstacles, my team has developed an apparatus that allows us to measure contact events – with a sensor sheet covering the object’s surface – and track time-varying hand postures – using deep learning-based computer vision – with unprecedented precision as animals interact with objects. We then characterize the responses at every stage along the somatosensory neuraxis, from peripheral nerve through cortex. This novel experimental set up will allow us to study the neural basis of somatosensation – particularly as it relates to manual dexterity – under ecologically valid conditions. In a related line of inquiry, we leverage what we learn about sensory processing to restore the sense of touch to bionic hands. In brief, we develop algorithms to convert the output of sensors on the bionic hand into patterns of electrical stimulation of the peripheral nerve (for amputees) or of somatosensory cortex (for people with tetraplegia) to evoke meaningful tactile percepts. I am one of the principal architects of the biomimetic approach to artificial touch, which posits that encoding algorithms that mimic natural neural signals will give rise to more intuitive tactile percepts, thereby endowing bionic hands with greater dexterity. Our work on artificial touch comprises three components: evaluation of the perceptual correlates of electrical stimulation, development of sensory encoding algorithms, and assessment of the benefits of artificial touch to manual behavior. The interplay of the basic scientific results and neural engineering efforts will result in more naturalistic artificial touch for brain- controlled bionic hands.