Advanced machine learning techniques for analysis of nanoscopic images

用于分析纳米级图像的先进机器学习技术

• 批准号: 2636081
• 金额: --
• 依托单位: University of Liverpool
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2021
• 资助国家: 英国
• 起止时间: 2021 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2636081
• 关键词: Advanced; machine; learning; techniques; analysis

This project has been developed by the University of Liverpool in partnership with Sivananthan Labaratories.Extracting quantitative information from nanoscale images of materials structures and processes is a repetitive and costly process. Currently humans are required to manually acquire images from experimentation and compare them with large repositories of data to find physical/chemical properties of nanoscale materials systems. In many cases, these databases often contain images with large disparities in their noise patterns, spatial/temporal resolution, magnification and/or signal-to-noise ratio, making the identification and use of these features challenging to accomplish in practice.In this project, our goal is to develop an optimised workflow for extracting quantitative information that is focused on one type of nanoscale image obtained by (scanning) transmission electron microscopy/(S)TEM. In many (S)TEM images, the size and shape (morphology) of nanoparticles and their special arrangements (dispersion) are a key analytical measure that has been demonstrated. Recent work by Chiwoo Park and Yung Ding has made significant strides in analysis of these types of images through implementation of existing algorithms to (S)TEM images. Their recent research outlines statistical techniques for the analysis of the morphology, location, dispersion analysis and multi-object tracking analysis of nanoparticles. These techniques can be leveraged to process existing unannotated data, and will serve as a baseline for performance for comparison with future models.This PhD will investigate generalised advanced analysis techniques to advance the capabilities (S)TEM imaging. Where applicable, supervised machine learning methods such as ANN, CNNs and Adversarial networks can be investigated for their ability to analyse images with varying degrees of noise within a microscopic setting. The aim of this project will be to leverage the support provided by the CDT to develop a generalised machine learning model capable of analysing hundreds of images. If successful, this will be used real-time during nanoscopy experiments, advancing the existing capabilities of (S)TEM research.

该项目由利物浦大学与Sivananthan实验室合作开发。从材料结构和过程的纳米级图像中提取定量信息是一个重复和昂贵的过程。目前，人类需要手动从实验中获取图像，并将其与大型数据库进行比较，以找到纳米材料系统的物理/化学特性。在许多情况下，这些数据库通常包含在其噪声模式、空间/时间分辨率、放大率和/或信噪比方面具有较大差异的图像，使得这些特征的识别和使用在实践中具有挑战性。我们的目标是开发一种优化的工作流程，用于提取定量信息，该信息集中在一种类型的纳米级图像上，（扫描）透射电子显微镜/（S）TEM。在许多（S）TEM图像中，纳米颗粒的大小和形状（形态）及其特殊排列（分散）是已被证明的关键分析指标。Chiwoo Park和Yung Ding最近的工作通过将现有算法应用于（S）TEM图像，在分析这些类型的图像方面取得了重大进展。他们最近的研究概述了用于分析纳米颗粒的形态，位置，分散分析和多目标跟踪分析的统计技术。这些技术可以用来处理现有的未注释的数据，并将作为与未来模型进行比较的性能基线。该博士将研究通用的高级分析技术，以提高TEM成像的能力。在适用的情况下，可以研究监督机器学习方法（如ANN，CNN和对抗网络）在微观环境中分析具有不同程度噪声的图像的能力。该项目的目的是利用CDT提供的支持，开发一个能够分析数百张图像的通用机器学习模型。如果成功，这将在纳米实验中实时使用，提高（S）TEM研究的现有能力。