A Computational Framework for Distributed Registration of Massive Neuroscience Images

项目摘要 神经科学站在了对大脑如何工作的新深度理解的边缘，这要归功于最近的 成像数据采集技术的进展，如光片fl荧光显微镜。怎么- 从来没有分析工具来挖掘这些丰富信息之间的跨样本、跨时间点和跨数据的关系 获取技术阻止研究人员解锁数量关系。我们建议创造 一种易于使用的分布式计算图像配准工具，它将把大图像映射到普通的 参考坐标系。这项工作将基于开源的Insight工具包(ITK)，这是一个得到广泛支持的标准 用于可重现的计算图像分析的库。我们建议通过以下方式扩展ITK的注册体系结构 来自深度学习和科学fic++社区的技术和方法，以有效地注册LSFM 数量和时间序列。该项目有可能将细胞分类和电路方面的最新进展结合在一起 这将最终阐明大脑发育和功能的潜在机制。

Project Summary Neuroscience stands at the precipice of a new depth of understanding about how the brain works thanks to recent advances in imaging data acquisition technologies such as light-sheet fluorescence microscopy (LSFM). How- ever, the lack of analytic tooling to mine this rich information's relationship across samples, timepoints, and data acquisition technologies prevents researchers from unlocking quantitative relationships. We propose the creation of an easy-to-use, distributed-computation image registration tools that will map large images into a common reference frame. This work will be based on the open source Insight Toolkit (ITK), a widely supported, standard library for reproducible, computational image analysis. We propose extending ITK's registration architecture with technologies and methods from deep learning and the scientific Python community to effectively register LSFM volumes and time series. This project has the potential to integrate recent advances in cell typing and circuit mapping that will ultimately elucidate the underlying mechanisms of brain development and function.