Per Anders Kaestner: Catalogue data in Spring Semester 2021
|Name||Mr Per Anders Kaestner|
Paul Scherrer Institut (PSI)
5232 Villigen PSI
|Department||Information Technology and Electrical Engineering|
|227-0966-00L||Quantitative Big Imaging: From Images to Statistics||4 credits||2V + 1U||P. A. Kaestner, M. Stampanoni|
|Abstract||The lecture focuses on the challenging task of extracting robust, quantitative metrics from imaging data and is intended to bridge the gap between pure signal processing and the experimental science of imaging. The course will focus on techniques, scalability, and science-driven analysis.|
|Objective||1. Introduction of applied image processing for research science covering basic image processing, quantitative methods, and statistics.|
2. Understanding of imaging as a means to accomplish a scientific goal.
3. Ability to apply quantitative methods to complex 3D data to determine the validity of a hypothesis
|Content||Imaging is a well established field and is rapidly growing as technological improvements push the limits of resolution in space, time, material and functional sensitivity. These improvements have meant bigger, more diverse datasets being acquired at an ever increasing rate. With methods varying from focused ion beams to X-rays to magnetic resonance, the sources for these images are exceptionally heterogeneous; however, the tools and techniques for processing these images and transforming them into quantitative, biologically or materially meaningful information are similar. |
The course consists of equal parts theory and practical analysis of first synthetic and then real imaging datasets. Basic aspects of image processing are covered such as filtering, thresholding, and morphology. From these concepts a series of tools will be developed for analyzing arbitrary images in a very generic manner. Specifically a series of methods will be covered, e.g. characterizing shape, thickness, tortuosity, alignment, and spatial distribution of material features like pores. From these metrics the statistics aspect of the course will be developed where reproducibility, robustness, and sensitivity will be investigated in order to accurately determine the precision and accuracy of these quantitative measurements. A major emphasis of the course will be scalability and the tools of the 'Big Data' trend will be discussed and how cluster, cloud, and new high-performance large dataset techniques can be applied to analyze imaging datasets. In addition, given the importance of multi-scale systems, a data-management and analysis approach based on modern databases will be presented for storing complex hierarchical information in a flexible manner. Finally as a concluding project the students will apply the learned methods on real experimental data from the latest 3D experiments taken from either their own work / research or partnered with an experimental imaging group.
The course provides the necessary background to perform the quantitative evaluation of complicated 3D imaging data in a minimally subjective or arbitrary manner to answer questions coming from the fields of physics, biology, medicine, material science, and paleontology.
|Lecture notes||Available online. https://imaginglectures.github.io/Quantitative-Big-Imaging-2021/weeklyplan.html|
|Literature||Will be indicated during the lecture.|
|Prerequisites / Notice||Ideally, students will have some familiarity with basic manipulation and programming in languages like Python, Matlab, or R. Interested students who are worried about their skill level in this regard are encouraged to contact Anders Kaestner directly (email@example.com).|
More advanced students who are familiar with Python, C++, (or in some cases Java) will have to opportunity to develop more of their own tools.