CAREER: Modeling Personalized Brain Development with Big Data

大数据提供了一个机会来研究特定的对照人群(年龄/性别/环境因素/人口统计学/遗传学)，并确定实质性的同质性亚群，以便人们可以理解潜在因素在大脑发育中所起的作用，区分异常轨迹和正常发育。图像处理、统计和信息学工具还不存在，可以有效和高效地利用大数据成像档案进行定量的人口水平研究和个性化医学。这项研究将创造新的信息学资源，将3D图像档案与可访问的研究数据库联系起来，从而使发现科学的规模比传统研究设计的常规可能要大得多。这项研究将发现影响个人大脑发育的遗传和环境因素，并通过个人发展轨迹来描述发育中的人脑。为了实现这一目标，将创建新的信息学技术，以实现(1)在异质、低质量和容易出错的数据环境中基于图像内容的图像处理和分割，以及(2)大型医学成像数据集的常规归档、查询和图像处理。这项研究将通过新的计算模型影响(1)信息学，(2)通过新的大脑发育结构模型影响神经科学，(3)通过新的可获得的研究数据集影响公共健康。利用大数据了解个性化大脑发育所带来的科技创新，将以分层的方式进行沟通。面向K-12的受众将致力于概念化设计标准，通过互动演示启发学生，并为学生提供在实践工程项目中应用关键概念的能力。对于高级学生和研究人员，将开发新的可访问的课程材料和在线模块，以便其他人可以在本研究建立的基础上建立基础。新的软件、数据辩论工具和资源将通过围绕新的试验台基础设施组织的两个研究推动力创建，并在第三个教育/外展推力中综合。推力1(个人大脑轨迹)将专注于在大规模执行时从医学图像中提取有意义的信息，方法是(1)创建对图像质量、采集和传输错误的变化稳健的自动化方法，以及(2)在规模上实现有效的人在环控制。这项研究将扩展新的统计模型用于图像内容标记，同时采用工业工程的质量控制技术。通过将医学成像标准与源自社交网络和电子商务社区的大数据架构相集成，推力2(新型存储和处理)将创建新的医学成像数据模型，以描述数据获取/检索、存储、清理、访问/安全、查询和处理。该基础设施将提供对PB级图像档案的实际访问，与现有数据工作流集成，并与商用硬件有效配合运行。PI将开发和发布一个参考试验台，在考虑年龄、性别和人口统计学的同时，在计算机辅助检测(CADE)大脑异常的背景下评估新技术。使用试验台，研究人员和学生将能够有效地评估现有的和新兴的图像处理软件，以筛选潜在的预后标志。在STRUCT 3(教育和外展)中，研究成果将被整合到两个针对本科生的班级中，并通过既定的研究生/教师培训计划创建和发布互动在线模块。每年夏天，一名本科生和高中生将通过在互动演示平台内实施和扩展研究贡献来参与研究。在第二到第五个暑假，一名高中教师将利用演示平台协助开发针对高中生的课程。高中学生和教师将从纳什维尔地铁学校招聘，这些学校的少数民族人数较少，午餐费用较低。这些努力将创建一个开放源代码、开放硬件的系统，用于公开演示和K-12课堂练习。

Big data offer an opportunity to study specific control populations (age / sex / environmental factors / demographics / genetics) and identify substantive homogeneous sub-cohorts so that one may understand the roles that potential factors play in brain development, differentiating abnormal trajectories from normal development. The image processing, statistical, and informatics tools to effectively and efficiently use big data imaging archives for quantitative population-level research and personalized medicine do not yet exist. This research will enable discovery science on a scale considerably larger than routinely possible with traditional study designs by creating novel informatics resources that tie archives of 3-D images into accessible research databases. This research will discover genetic and environmental factors that influence an individual's brain development and characterize the developing human brain through personal developmental trajectories. To accomplish this goal, new informatics technologies will be created to enable (1) image processing and segmentation based on image content in the context of heterogeneous, low quality, and error prone data with minimal human oversight and (2) routine archival, query, and image processing of large medical imaging datasets. This research will impact the areas of (1) informatics via novel computation models, (2) neuroscience via a new structural model of brain development, and (3) public health via newly accessible data sets for research. The science and technology innovations enabled by using big data to understand personalized brain development will be communicated in a tiered method. Outreach to the K-12 audience will target conceptualizing design criteria, inspiring students with interactive demonstrations, and providing capabilities for students to apply key concepts in hands-on engineering projects. For advanced students and researchers, new accessible course materials and online modules will be developed so that others may build upon the foundations established by this research.Novel software, data wrangling tools, and resources will be created through two research thrusts organized around a novel test bed infrastructure and synthesized in a third education/outreach thrust. Thrust 1 (Personal Brain Trajectories) will focus on extracting meaningful information from medical images when performed at scale through (1) creating automated methods robust to variations in image quality, acquisition, and transfer errors, and (2) enabling efficient human-in-loop control at scale. The research will extend novel statistical models for image content labeling while adapting quality control techniques from industrial engineering. Thrust 2 (Novel Storage & Processing) will create novel medical imaging data models to describe data acquisition / retrieval, storage, cleaning, access / security, query and processing by integrating of medical imaging standards with big data architecture derived from social network and e-commerce communities. This infrastructure will provide practical access to petabyte imaging archives, integrate with existing data workflows, and effectively function with commodity hardware. The PI will develop and release a reference test bed to evaluate new technologies in the context of computer-aided detection (CADe) of brain abnormalities while considering age, sex, and demographics. Using the test bed, researchers and students will be able to efficiently evaluate existing and emerging image processing software to screen for potential prognostic markers. In Thrust 3 (Education and Outreach), the research results will be integrated into two classes targeting undergraduate students and interactive online modules created and released through an established graduate student/faculty training program. Each summer, an undergraduate and high school student will participate in research by implementing and extending research contributions within an interactive demonstration platform. In the second through fifth summers, a high school teacher will assist in the development of curricula targeting high school students using the demonstration platform. High school students and teachers will be recruited from Nashville Metro schools with a high underrepresented minority / reduced cost lunch populations. These efforts will create an open-source, open-hardware system for public demonstration and K-12 classroom exercises.