Automating Biomedical Data Analysis

该项目将创建自动化系统来分析大型多模式生物医学数据，以增强 教育和研究基础设施在肯特州立大学和超越。完成的任务将 为至少50个研究实验室提供支持并提高分析数据的能力（准确性和速度）， 并通过融入研究项目和课程培训了200多名学生。我们将（1） 创建算法来自动处理大型空间生物医学数据， 分析数千个细胞，2）创建自动处理动态磁共振的方法 成像（MRI）数据，3）在疾病的动物模型中经验性地评估新方法，以及，4） 生成与疾病中细胞群变化相关的数据，以提供新的治疗途径。作为 为了实现这些目标，我们将至少在三个方面支持和加强教育：1）加强 在三个实验室中进行生物医学成像研究技术的学生培训，2）创建重复的多个 围绕资源开发的学科课程，包括“应用生物医学数据 处理”和“生物图像分析”，以及，3）开发基础设施，以供继续使用， 为最终用户开发基于集群的并行数据处理系统，供学生和 全世界的研究人员。激光扫描和三维电子显微镜产生的数据 由成千上万的连续图像组成，构成大量数据。功能和结构 MRI系统通常用于使用多种模态在多个月内扫描受试者和患者 (fMRI弥散加权，T1/T2）。这些类型的阵列可以具有数千个图像和/或离散图像。 每个模态的时间点生成复杂数据，需要大量人工时间（天）来处理， 经常需要二次抽样。我们的长期目标是支持所有类型的自动化数据分析 与人类健康相关，因此该项目的主要重点是创建一个可扩展的平台， 方法来完全支持所有计算昂贵的数据分析。我们将首先集中精力 创建用于从大型显微镜数据（大量）中自动提取和分析神经胶质细胞的工具 空间组织图）、自动分割和分析动态MRI数据以及自动化大数据 使用并行系统进行处理。这些方法将用于评价细胞群的变化 在疾病的动物模型中发生，以确定新的治疗策略和操作。这 将大大加强肯特州立大学的研究和教育基础设施，包括 开发新的方法来自动化生物医学数据分析以及资源， 许多研究小组自动应用并继续使用这些程序和现有程序。此外，本发明还 通过创建新的课程，并在不同的实验室多学科研究活动的整合， 数百名学生将接受有关方法和技术的应用和发展方面的培训。

This project will create automated systems for analyzing big multimodal biomedical data to enhance the educational and research infrastructure at Kent State University and beyond. The completed tasks will provide support and improve the ability to analyze data (accuracy and speed) for at least 50 research labs, as well as train well over 200 students via integration into research programs and the curriculum. We will 1) Create algorithms to automate processing of large spatial biomedical data to automatically extract and analyze thousands of cells, 2) Create methods to automate the processing of dynamic magnetic resonance imaging (MRI) data, 3) Empirically evaluate the new methods in an animal model of disease, and,4) Generate data relating to changes in cell populations in disease to provide new therapeutic avenues. As we accomplish these goals we will support and strengthen education in at least three areas; 1) Enhance student training in biomedical imaging research techniques in three labs, 2) Create recurring multi- disciplinary courses based around development of the resources including “Applied Biomedical Data Processing” and “Biological Image Analysis” , and, 3) Develop the infrastructure for continued use and development for end users with a cluster-based parallelized data processing system for students and researchers worldwide. Laser scanning and three-dimensional electron microscopy produce data consisting of thousands of sequential images making up large volumes of data. Functional and structural MRI systems are routinely used to scan subjects and patients over many months using multiple modalities (fMRI, diffusion weighted, T1/T2). These types of arrays can have thousands of images and/or discrete time-points per modality generating complex data requiring significant human time (days) to process where sub-sampling is frequently required. Our long term goal is to support all types of automated data analysis pertinent to human health, and as such a major focus of this project is to create an extensible platform and methods to fully support all computationally expensive data analysis. We will initially focus our efforts on creating tools for the automated extraction and analysis of glial cells from large microscopy data (massive spatial tissue maps), automate segmentation and analysis of dynamic MRI data and automate big data processing using parallel systems. The methods will be used to evaluate changes to cell populations occurring in an animal model of disease to identify new strategies and manipulations for treatments. This will significantly enhance the research and educational infrastructure at Kent State University and include the development of new methods to automate biomedical data analysis as well as resources for the automated application and continued use of these and existing routines by many research groups. Further, by the creation of new courses, and integration of the multidisciplinary research activities in diverse labs, hundreds of students will be trained in the application and development of the methods and techniques.