CRCNS: Dense longitudinal neuroimaging to evaluate learning in childhood

了解儿童早期学习是如何发生的，有可能改变我们对 人类学习和我们构建智能机器的方法，但幼儿期的关键窗口 保持采样不足，因此提供关于学习的很少见解。一个基本的， 人类学习中一个长期存在的问题是， 数字处理在一年级开始。这些知识对于解决公共卫生问题至关重要 关注与阅读和数学素养，因为一年级的字母和数字知识是 未来阅读和数学能力的最强预测因子，以及在阅读和数学方面落后的儿童 在小学可能会经历医疗和财政不稳定的成年人。该项目采用 一个多层次的方法来理解儿童的学习，这将支持关键的进步， 几个学科，包括人类和人工学习，发展和认知神经科学， 教育神经科学、神经成像方法、计算机视觉和广泛的学习科学。的 第一个目标是创建和分发一个大型的图像语料库，从芝麻街插曲注释， 教育内容，如字母和数字，以及其他常见的对象类别。图像 语料库将是第一个捕捉儿童学习者的视觉统计数据，并可用于训练不同的 人工学习架构，以更好地理解人类学习。第二个目标是收集， 预处理，并分配在1000000处采样的大脑结构和功能的密集纵向MRI数据集。 在一年级的多个时间点密集纵向MRI数据集将提供 实验测量的大脑对芝麻街语料库图像的反应， 用于理解人类学习，并具有适当的规模以约束人工学习架构。 第三个目的是评估字母和数字作为学习的选择性神经处理的出现 在整个学校的第一年。这一目标将解决人类学习中的一个悬而未决的问题 关于字母和数字的神经特化出现的过程，即 运动系统在新兴的专业化。了解大脑功能变化的时间过程 在早期学习过程中的认知和结构对于发展长期生活结果的准确预测至关重要 以及用于识别具有很大可塑性的敏感窗口以优化干预时间表。

Understanding how learning occurs in early childhood has the potential to transform our understanding of human learning and our approach to building intelligent machines, yet critical windows in early childhood remain under-sampled and consequently provide little insight concerning learning. One fundamental and long-standing question in human learning is the process by which neural specialization for visual letter and digit processing emerges in the first grade. This knowledge is critical for addressing public health concerns related to reading and math literacy because first-grade letter and digit knowledge are the strongest predictors of future reading and math abilities, and children who fall behind in reading and math in elementary school will likely experience medical and financial instability as adults. This project employs a multi-level approach to understanding learning in childhood that will support critical advancements in several disciplines, including human and artificial learning, developmental and cognitive neuroscience, educational neuroscience, neuroimaging methods, computer vision, and learning sciences broadly. The first aim is to create and distribute a large corpus of images from Sesame Street episodes annotated for educational content, such as letters and digits, as well as for other common object categories. The image corpus will be the first to capture the visual statistics of child learners and can be used to train different artificial learning architectures to better understand human learning. The second aim is to collect, preprocess, and distribute a dense longitudinal MRI dataset of brain structure and function sampled at multiple time points throughout the first grade year. The dense longitudinal MRI dataset will provide experimentally measured brain responses to images from the Sesame Street corpus that will be of benefit for understanding human learning and of appropriate scale for constraining artificial learning architectures. The third aim is to evaluate the emergence of selective neural processing for letters and digits as learning occurs throughout the first year of schooling. This aim will address an open question in human learning concerning the process by which neural specialization for letters and digits emerges, namely the role of the motor system in emerging specialization. Understanding the time course of changes in brain function and structure during early learning is critical for developing accurate predictors of long-term life outcomes and for identifying sensitive windows of great plasticity to optimize intervention timelines.