BRAIN EAGER:The Virtual Neuroanatomist: Using Machine Intelligence to Study Intelligent Machines

BRAIN EAGER：虚拟神经解剖学家：利用机器智能研究智能机器

• 批准号: 1450957
• 负责人: Partha Mitra
• 金额: $ 30万
• 依托单位: Cold Spring Harbor Laboratory
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2014
• 资助国家: 美国
• 起止时间: 2014-09-01 至 2018-02-28
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1450957&HistoricalAwards=false
• 关键词: BRAIN; EAGER; Virtual; Neuroanatomist; Using

Cutting-edge light microscopy technology allows entire vertebrate brains to be digitized, resulting in data sets of unprecedented size and complexity. However there is a lack of adequate computational tools to visualize, manage, analyze, and disseminate these enormous data sets, and visual examination by expert human neuroanatomists remains the standard method to extract information from microscopic images. Software tools exist that can perform simple operations, but they are not able to adequately mimic the visual pattern recognition skills of an experienced neuroanatomist. This project aims to develop computational tools that mimic the analysis of an expert neuroanatomist, thus allowing for rich data analysis of whole-brain light microscopy data sets on a scale that has been previously intractable using human experts.Machine vision algorithms will be developed and integrated into an an open source software toolbox (with associated whole brain image data) that will be made widely accessible for further development and refinement. Pattern recognition methodology will be applied to combine information about brain location (e.g., "where are we in the brain") with information about correspondence of brain structures in different species (e.g., "which areas of the brain of these species correspond"). Incorporating comparative neuroanatomical knowledge into pattern-recognition methodology is radically different from the "atlas morphing" approach currently used, and has the potential to transform the study of whole-brain neuroanatomy. The tools will help fill a gap in knowledge and skills (as contemporary neuroanatomists are increasingly less frequently being trained to study whole-brain microscopic anatomy), and postdoc and PhD students will be trained in the project.

尖端的光学显微镜技术使整个脊椎动物的大脑数字化，从而产生前所未有的规模和复杂性的数据集。然而，缺乏足够的计算工具来可视化，管理，分析和传播这些巨大的数据集，由人类神经解剖学家专家进行的视觉检查仍然是从显微图像中提取信息的标准方法。软件工具存在，可以执行简单的操作，但它们不能充分模仿有经验的神经解剖学家的视觉模式识别技能。该项目旨在开发模拟专家神经解剖学家分析的计算工具，从而允许对全脑光学显微镜数据集进行丰富的数据分析，其规模在以前是人类专家难以处理的。机器视觉算法将被开发并集成到一个开源软件工具箱中。（与相关的全脑图像数据），其将被广泛访问以用于进一步开发和改进。模式识别方法将被应用于联合收割机关于大脑位置的信息（例如，“我们在大脑中的位置”）与关于不同物种中大脑结构的对应性的信息（例如，“这些物种的大脑区域对应”）。将比较神经解剖学知识转化为模式识别方法与目前使用的“图谱变形”方法完全不同，并有可能改变全脑神经解剖学的研究。这些工具将有助于填补知识和技能方面的空白（因为当代神经解剖学家越来越少接受全脑显微解剖学研究的培训），博士后和博士生将在该项目中接受培训。