Computational Image Analysis for Cellular and Developmental Biology

描述(由申请者提供)：本提案请求支持一个为期十天的细胞和发育生物学计算图像分析强化课程。这门课程是为研究生和博士后研究员设计的，在马萨诸塞州伍兹霍尔的海洋生物实验室进行。该课程是此类课程中的第一门，为学生提供计算机视觉方面的正式培训，以便对细胞和发育生物学图像数据进行具体分析。在这一主题中建立坚实的基础对于推动细胞和发育生物学向前发展至关重要，因为成像已经成为这些领域中越来越不可或缺的工具。本课程涵盖计算机视觉的基本原理，带领学生完成解决定量细胞和发育生物学中的图像分析问题所需的低级、中级和高级计算机视觉任务序列。课程从滤波、阈值和边缘/线/通用特征检测开始，然后是使用模型拟合的更复杂的检测算法。在介绍了低级计算机视觉任务之后，本课程继续进行中级和高级任务，包括对象在空间和时间上的关联(如跟踪)和用于表型分类的机器学习工具。每个主题首先由四名核心教员之一讲授，然后是3-4个小时的计算机编程课程，学生们立即将他们学到的概念付诸实践。通常一天有两堂课，计算机实验室。大多数编程练习都是单独的，让每个学生都有机会“动手”，还有两个是团队项目，允许学生学习和实践代码共享的方法。在为期十天的课程中，有三场由生物成像和计算机视觉领域的顶尖研究人员主持的客座讲座，每一场讲座之后都会有深入的讨论，以及学生、教职员工和助教的研究讲座。在此基础上，核心讲座和实验以逻辑连续的方式向学生传授计算机视觉的基础知识，客座讲师向学生介绍成像和图像分析方面令人兴奋的新挑战，学生/教师研究讲座鼓励所有课程参与者之间的交流，特别是让学生有机会反思课程如何帮助他们进行研究。

DESCRIPTION (provided by applicant): This proposal requests support for an intensive ten-day course on Computational Image Analysis for Cellular and Developmental Biology. The course is designed for graduate students and postdoctoral fellows, and takes place at the Marine Biological Laboratory in Woods Hole, MA. The course is the first of its kind, giving students formal training in computer vision for the specific analysis of cell and developmental biology image data. Building strong foundations in this topic is critical for pushing cell and developmental biology forward, as imaging has become more and more an indispensable tool in these fields. The course covers the fundamentals of computer vision, taking the students through the sequence of low-, intermediate-, and high-level computer vision tasks that are required to solve image analysis problems in quantitative cellular and developmental biology. The curriculum starts with filtering, thresholding and edge/line/generic feature detection, followed by more sophisticated detection algorithms that employ model fitting. After this introductory block to low-level computer vision tasks, the course moves on to intermediate and higher-level tasks, including object association in space and time (such as tracking) and machine learning tools for phenotype classification. Each topic is covered first by a lecture, generally taught by one of the four core faculty, followed by a 3-4 hour computer programming session where students immediately implement the concepts they learn. There are usually two lectures + computer labs per day. Most programming exercises are individual, giving each student the opportunity to "get their hands dirty," while two are team projects allowing the students to also learn and practice methods of code sharing. Over the course's ten days there are three guest lectures by leading researchers in the fields of biological imaging and computer vision, each followed by in-depth discussion, as well as research talks given by the students, faculty and teaching assistants. With this, the core lectures and labs teach the students the fundamentals of computer vision in a logical, continuous manner, the guest lecturers introduce the students to exciting new challenges in imaging and image analysis, and the student/faculty research talks encourage communication between all course participants and give especially the students the opportunity to reflect on how the course can help them with their research.