Collaborative Research: NCS-FO: A model-based approach to probe the role of spontaneous movements during decision-making

认知任务，如决策，是由不同的主体以不同的方式执行的。受试者与受试者之间的变异性的一个重要表现是在任务执行过程中的自发运动。例如，人们在学习一项认知任务时可能会敲打地板或移动眼睛；类似地，老鼠会表现出特殊的胡须、面部或其他动作。由于自发运动和认知任务都调节大脑皮层活动，对决策过程中的神经活动进行建模和解释是一个重大挑战。在明确建模个体神经动力学的同时，破译自发运动的功能并量化其神经起源，是理解认知行为的神经基础的关键第一步。阐明自发运动在认知中的作用对于评估和开发治疗注意缺陷多动障碍等神经行为障碍的新疗法也很重要。由于认知任务中的神经活动受到运动的强烈调制，并且自发的、无指导的运动对学习和认知的影响还不清楚，因此无法完全理解认知任务中的大规模神经和行为数据。将与运动相关的神经动力学与与认知任务相关的神经动力学区分开来的一个重大挑战是存在受试者对受试者的可变性。传统上，可变性是通过关注存在于大量主题中的表示来解决的。相比之下，这项研究将通过将自发运动和受试者对受试者的可变性与任务的跨受试者部分分离来明确地对其进行建模。具体地说，这个项目将通过学习独立于受试者的动力学模型，同时单独量化受试者特定的动力学，来利用小鼠神经活动中的跨受试者相似性。此外，下一代神经刺激算法不应该要求在每个主题上都进行广泛的培训。为了实现这一目标，该项目将使用一个训练有素的跨学科模型作为新学科的初始模型，同时用个别数据进一步完善它。这种迁移学习方法将导致有效的基于模型的控制策略来设计闭环式神经刺激。在一种综合的计算和实验方法中，这些模型将用于调查高维皮质活动对自发运动和认知任务的贡献。这一奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Cognitive tasks such as decision-making are performed by different subjects in different ways. An important manifestation of this subject-to-subject variability is through spontaneous movements during task performance. For example, people may tap the floor or move their eyes while learning a cognitive task; similarly, mice display idiosyncratic whisker, facial, or other movements. Since both spontaneous movements and cognitive tasks modulate cortical activity, modeling and interpreting neural activity during decision-making is a major challenge. Deciphering the function and quantifying the neural origin of spontaneous movements while explicitly modeling individual neural dynamics is a crucial first step in understanding the neural basis of cognitive behaviors. Elucidating the role of spontaneous movements in cognition will also be important for assessing and developing novel therapies for neurobehavioral disorders such as attention deficit hyperactivity disorder.It has not been possible to fully understand large-scale neural and behavioral data during cognitive tasks because neural activity is strongly modulated by movements during the task, and the effect of spontaneous, uninstructed movements on learning and cognition is not well understood. A significant challenge in disentangling the neural dynamics related to movements from those related to cognitive tasks is the presence of subject-to-subject variability. Traditionally, the variability is resolved by focusing on the representations that are present across a large number of subjects. In contrast, this study will explicitly model spontaneous movements and subject-to-subject variability by decoupling it from the across-subject part of the task. Specifically, this project will leverage across-subject similarity in neural activity in mice by learning subject-independent dynamical models, while separately quantifying subject-specific dynamics. Furthermore, the next generation of neurostimulation algorithms should not require extensive training on each subject. For this goal, this project will use a trained across-subject model as the initial model for new subjects, while refining it further with individual data. This transfer learning approach will lead to effective model-based control strategies to design closed-loop neurostimulation. In an integrative computational and experimental approach, these models will be used to investigate the contribution of high-dimensional cortical activity to spontaneous movements and cognitive tasks.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.