Developing an Adaptive Cognitive Prosthetic to Replace Damaged Brain Regions

描述（由申请人提供）：创伤性脑损伤严重降低了受影响个体的生活质量，并带来重大的社会和经济成本。尽管这些损伤普遍存在，但目前没有有效的治疗方法。在这里，我们建议开发一种新的治疗永久性脑损伤的方法：一种自适应认知假体，它将学会取代由于脑损伤而失去的神经功能。为了有效地绕过受损的皮层区域，我们提出的认知假体将模仿其结构：它将记录来自其他大脑区域的神经活动，根据失去的认知功能转换这种活动，然后刺激未受影响的大脑区域以传达结果。在这里，我们建议开发实现这样一个假体所需的三个核心组件。首先，神经活动必须从未受影响的大脑区域记录下来。这将为假肢提供输入，使其能够监控当前的大脑状态。第二，假肢 必须能够作用于大脑，刺激未受损大脑区域的神经组织，以绕过受损的大脑区域。最后，自适应算法必须将前两个部分连接起来：将观察到的大脑状态转换为刺激模式，以模拟将要被替换的大脑区域。然而，先前与每个状态相关联的刺激的确切模式是不可知的，因此必须学习。为此，我们提出了一种新的，分层的，学习算法，可以发现适当的刺激模式。我的实验室在开发这种假肢方面有得天独厚的优势。我们有大量的神经元记录的丰富经验（第一部分），并开发了一种新的范式来刺激神经活动的模式（第二部分）。在当前的提案中，我们将开发自适应算法，在小鼠中测试其功效。最后，我们将联合收割机的所有必要的组成部分，以测试认知假肢在猴子模型的半侧空间忽视，一个共同的行为缺陷后，顶叶中风。

DESCRIPTION (provided by applicant): Traumatic brain injuries severely reduce the quality of life for affected individuals and carry significant societal and economic costs. Despite the prevalence of these injuries, there is currently no effective treatment. Here we propose to develop a novel treatment for permanent brain damage: an Adaptive Cognitive Prosthetic that will learn to replace the neural function that was lost due to a brain injury. To effectively bypas the damaged cortical region, our proposed cognitive prosthetic will mimic its structure: it will record neural activity from other brain regions, transform this activity according to the lost cognitive function, and then stimulate unaffected brain regions in order to convey the result. Here we propose to develop the three core components necessary for achieving such a prosthetic. First, neural activity must be recorded from unaffected brain regions. This will provid the input to the prosthetic, allowing it to monitor the current brain state. Second, the prosthetic must be able to act upon the brain, stimulating neural tissue in undamaged brain regions in order to bypass the damaged brain region. Finally, an adaptive algorithm must bridge the previous two components: transforming the observed brain state to a pattern of stimulation in order to mimic the to-be-replaced brain region. However, the exact pattern of stimulation previously associated with each state is unknowable and therefore must be learned. To this end, we propose a novel, hierarchical, learning algorithm that can discover the appropriate stimulation patterns. My lab is uniquely positioned to develop this prosthetic. We have extensive experience recording from large populations of neurons (the first component) and have developed a novel paradigm for stimulating patterns of neural activity (the second component). In the current proposal we will develop the adaptive algorithm, testing its efficacy in mice. Finally, we will combine all of the necessary components to test the cognitive prosthetic in a monkey-model of hemispatial neglect, a common behavioral deficit following parietal stroke.