Machine Vision-Based Quantification of Plant Growth and Development

PI: Edgar Spalding（威斯康辛大学）PI: Amir H. Assadi和Nicola J. Ferrier（威斯康辛大学）植物在自然界和社会中起着至关重要的作用，这就是为什么了解它们如何在深层次上起作用是现代生物学的主要目标。具体地说，植物生物学家正在努力了解植物DNA中存在的数千个基因中的每一个对植物的生长、发育、生理和生物化学有何贡献。虽然已知一些基因直接影响种子产量、抗逆性、开花时间和其他重要的农艺性状，但任何植物中的大多数基因都没有已知的功能。了解基因功能的一个非常有效的方法是确定当基因发生突变时，植物会发生什么变化。这种差异，或表型，通常提供了关于突变基因功能的线索。大量的突变植物已经在各种物种中产生，特别是水稻、玉米、紫花苜蓿（苜蓿的亲戚）和拟南芥（油菜的亲戚）。在许多情况下，研究人员没有在这些突变体中发现表型，可能是因为表型很微妙，或者因为它影响了植物的一个不显眼的部分，或者因为它是短暂表达的。目前的项目是一个跨学科努力开发机器视觉技术的试点版本，用于发现突变或自然变化的植物种群的表型。植物生物学家、工程师和数学家正在合作开发一个筛选平台，该平台采用电子CCD相机、机器人定位设备和定制计算工具，以量化和数学方式表征（分类）根、茎和叶等结构的生长和发育。在这个试验阶段，硬件、自定义算法和数据管理的最佳组合将被组装成一个平台，该平台将通过筛选一组拟南芥突变体和重组自交系的根系生长和向地性进行测试。结果将通过可搜索的数据库以标准化格式提交给社区，这些数据库通过特定项目的网站和TAIR与万维网相连。原始数据将提供给图像分析社区，以帮助他们开发新的算法和分类技术，促进更多的协作工具开发。根据试点阶段的成功，后续阶段将建立一个平台，用于系统地、高通量地发现和定量玉米、水稻、番茄和豆类等作物的表型。这个项目更广泛的影响将来自工程师、生物学家、计算机科学家和数学家的集合。这些人员将在独特的跨学科环境中接受培训，从而增加了这种跨领域研究成为常态而不是例外的可能性。为了在早期教育阶段增加计算与植物发育之间的联系，拓展活动将把图像分析引入高中课堂。正在开发中的植物结构的动态图像序列，以及如何创建能够从图像中提取信息的图像分析算法的提示，将通过一个特定项目的网站提供，以便高中计算机科学课程能够掌握这些原理，并了解计算机科学、工程学和植物生物学如何相互联系。随着该项目的发展，将开发一系列课堂上使用的生物图像/计算材料，包括pi在内的参与项目的人员将协助将其整合到K-12课程中。

PI: Edgar Spalding (University of Wisconsin)CoPIs: Amir H. Assadi and Nicola J. Ferrier (University of Wisconsin)Plants serve critical functions in nature and society, which is why understanding how they work at a deep level is a major goal in modern biology. Specifically, plant biologists are endeavoring to learn what each of the many thousands of genes present in a plant's DNA contributes to the growth, development, physiology, and biochemistry of the plant. While some genes are known to contribute directly to seed yield, stress resistance, flowering time, and other agronomically important traits, most genes in any plant have no known function. One very effective way to learn the function of a gene is to determine what happens differently in a plant when the gene is mutated. This difference, or phenotype, often provides a clue about the function of the mutated gene. Large numbers of mutant plants have been produced in a variety of species, particularly rice, corn, Medicago (a relative of alfalfa) and Arabidopsis (a relative of canola). In many cases, researchers have not found phenotypes in these mutants, possibly because the phenotype is subtle, or because it affects an inconspicuous part of the plant, or because it is transiently expressed. The present project is a pilot version of an interdisciplinary effort to develop machine-vision technology for discovering phenotypes in mutant or naturally varying populations of plants. Plant biologists, engineers, and mathematicians are collaboratively developing a screening platform that employs electronic CCD cameras, robotic positioning devices, and custom computational tools to quantify and mathematically characterize (classify) the growth and development of structures such as roots, stems, and leaves. During this pilot phase, the best combination of hardware, custom algorithms, and data management will be assembled into a platform that will be tested by screening root growth and gravitropism in a selected set of Arabidopsis mutants and recombinant inbred lines. The results will be presented to the community in a standardized format via searchable databases linked to the World Wide Web through a project-specific website and TAIR. The raw data will be available to the image-analysis community to assist their development of new algorithms and classification techniques, fostering more collaborative tool development. Depending on the success of the pilot phase, a subsequent phase will establish a platform for the systematic, high throughput discovery and quantification of phenotypes in crop plants such as maize and rice and tomato and legumes. Broader impacts of this project will come from the bringing together of engineers, biologists, computer scientists, and mathematicians. The personnel will be trained in a unique interdisciplinary environment, increasing the likelihood that such cross-cutting research becomes the norm rather than the exception. To increase the connections between computation and plant development at earlier educational stages, outreach activities will bring image-analysis to the high school classroom. Dynamic image sequences of plant structures undergoing development and tips on how to create image-analysis algorithms capable of extracting information from the images will be made available through a project-specific website so that high school computer science classes can grapple with the principles and see how computer science, engineering, and plant biology can interrelate. As this project develops, a host of classroom-ready bioimage/computation materials will be developed and personnel involved in the project including the PIs will assist in their integration into K-12 curricula.