Inference and Model Building for Vision-based Estimation of Transmissive Objects

本提案旨在进一步发展计算机视觉和图像处理领域的最新技术，目标是扩大和加强成像模式在我们对世界的理解中的使用。图像是丰富的信息来源，在许多场景中都很容易捕获。处理这些图像并从中提取意义的能力是计算机视觉的关键目标。该提案和PI研究计划的长期目标是在理论和实践上发展模型，使人们能够从图像中估计世界的属性。其中一个问题是对透射物体的详细三维结构的估计。这些物体的成像和重建出现在一系列领域，包括：云、玻璃、水或烟雾等半透明物体的光学成像，用于医疗目的的人类和动物的x射线成像，以及材料和生物标本的透射电子成像。本提案的目标是聚焦于透射电子成像，旨在建立电子低温显微镜（cryo-EM）的新方法。具体的问题有关的估计三维结构的生物分子，如蛋白质和病毒从冷冻电镜图像考虑。低温电镜是一种新兴的生物分子实验结构测定方法。DNA的功能有时被类比为生命的源代码。继续这个类比，蛋白质和病毒是经过编译的程序，它们采取分子机器的形式，在细胞水平上“执行生命”。确定这些机器的三维结构并了解它们是如何工作的，这不仅是基本的科学兴趣，而且对理解疾病和开发新药和治疗方法也至关重要。这里提出的研究将推进最先进的技术，并对冷冻电子显微镜产生重大影响，从而对加拿大和世界各地的科学和医学的许多方面产生重大影响。通过提高估计结构的可靠性和分辨率，本文提出的方法将为快速结构确定方法奠定基础，从而可以更快地开发新药，减少产生意想不到的副作用的可能性。它们还可以让科学家们研究以前从未见过的细胞机制，因为这些分子机器的运动可以被直接捕获。

This proposal seeks to further the state of the art in the field of computer vision and image processing, with the goal of expanding and reinforcing the use of imaging modalities in our understanding of the world. Images are a rich source of information which are readily captured in many scenarios. The ability to process these images and extract meaning from them is the key goal of computer vision. The long-term objective of this proposal and the PI's research program in general is the theoretical and practical development of models which enable the estimation of properties of the world from images. One such problem is the estimation of detailed 3D structure of transmissive objects. Imaging and reconstruction of such objects arises in a range of domains including: optical imaging of translucent objects like clouds, glass, water or smoke, x-ray imaging of humans and animals for medical purposes, and transmission electron imaging or materials and biological specimens.The objectives of this proposal focus on transmission electron imaging with the aim to build new methods for electron cryomicroscopy (cryo-EM). Specific problems relating to the estimation of the 3D structure of biological molecules such as proteins and viruses from cryo-EM images are considered. Cryo-EM is a rapidly emerging method for experimental structure determination of biological molecules. The function of DNA has sometimes been explained by analogy as the source code of life. Continuing this analogy, proteins and viruses are the compiled programs which take the form of molecular machines that “execute life” at a cellular level. Determining the 3D structures of these machines and understanding how they work is not only of fundamental scientific interest, it is also critical in the understanding of disease and the development of novel drugs and treatments.The research proposed here would advance the state-of-the-art and have significant impacts on cryo-EM and consequently on many aspects of science and medicine both within Canada and around the world. By improving the reliability and resolution of estimation structures, the methods proposed here would form the basis for rapid structure determination methods which could be used to more quickly develop new drugs which would be less likely to have unexpected side-effects. They would also allow scientists to study never before seen mechanisms in the cell as the motion of these molecular machines could be captured directly.