Undergraduate Course: Scientific Image Analysis (PHYS11071)
Course Outline
| School | School of Physics and Astronomy |
College | College of Science and Engineering |
| Credit level (Normal year taken) | SCQF Level 11 (Year 4 Undergraduate) |
Availability | Available to all students |
| SCQF Credits | 20 |
ECTS Credits | 10 |
| Summary | An introduction to scientific image analysis i.e. the extraction of quantitative information from scientific images. Both general principles and practical implementation will be covered, including machine-learning approaches. Examples will be drawn from scientific research, but relevance to applications beyond academia will also be covered. The course will also cover examples of scientific imaging techniques, as understanding contrast mechanisms is crucial when interpreting results from image analysis. Both an open-source application (e.g. ImageJ) and a contemporary programming language (e.g. Python) will be used as image-analysis tools, as both tools are widely used for preliminary and batch-processed image analysis.
The course starts with a description of images, image properties, and images as numerical arrays. Image acquisition, contrast mechanisms, and image artefacts will be covered. Several techniques for extracting quantitative information from images will be considered, including machine-learning approaches. Examples of applying image analysis to solving problems in physics and astronomy research, and beyond, will be presented. |
| Course description |
Image analysis involves the extraction of quantitative information from images, image sequences or videos; these days the images are usually digital. It is used in academic research, industry e.g. quality control in manufacturing processes, and medical research e.g. analyzing X-ray CT images. Within the School, it is used in Astronomy, (Soft) Condensed Matter and Particle Physics. Hence, it is a skill that is regularly encountered in both undergraduate and postgraduate (research) projects. Overall, it is a skill that may well be useful to any graduate, whether planning a career within or outwith academia.
Images are typically easy to record (e.g. with a mobile phone) but non-trivial to analyze. In this course, you will learn about scientific images and how to quantitatively analyze them, for example by writing a computer program in such a way that it reliably distinguishes the objects of interest from the background. This may sound trivial, but low signal-to-noise, motion blurring, non-uniform illumination, loss of contrast, etc can make it challenging to extract the desired information from the image. You will also consider image acquisition and how it may affect image analysis, as good image analysis starts with at least a basic understanding of the imaging method. This is because all imaging methods are prone to artefacts, which can lead to misunderstanding; you are more likely to identify artefacts if you understand the imaging method. Moreover, in many cases, the best way to improve quantitative analysis of images is to record higher-quality images.
This course has a substantial hands-on component i.e. a substantial amount of time is spent on directed independent learning, supported by lab sessions, drop-in classes and an online forum. Both ImageJ and Python will be used as image-analysis tools, as both tools are widely used for preliminary (ImageJ) and batch-processed (Python) image analysis. The lectures will mainly focus on general principles, for example thresholding considerations and errors introduced by thresholding. The course also covers topics beyond coding, including the physics of image contrast generation. The course is rooted in the application of image analysis to physics and astronomy research; examples could include: extracting particle size from microscopy images for scientific research and industrial quality control, quantitative analysis of 3D X-ray CT scans for materials characterization and medical applications, and extracting information from data where image quality may be affected by clouds and/or lighting, for example in astronomy and satellite data. Notably, these examples build on each other and are chosen to demonstrate the wide applicability of scientific image analysis, not because techniques are mutually exclusive. For example, image segmentation may be introduced in analyzing electron microscopy images, but it is also ubiquitous in analyzing images in medical applications and astronomy.
Topics covered in this course are:
- Images, image properties, images as numerical arrays, version control.
- Image acquisition, contrast mechanisms, sample preparation, image artefacts.
- Image processing, filtering, enhancement, Fast Fourier Transforms.
- Segmentation, morphological operations, measurements and errors, image analysis without thresholding.
- Image sequences, time and memory considerations, machine-learning approaches, training considerations.
- Applications of image analysis in scientific research and beyond.
The course will be taught through a combination of lectures, lab sessions, drop-in classes and an online discussion forum. The lectures only run in the first half of the course, to allow more time to be dedicated to the main assignment in the second half. Lab sessions and drop-in classes run throughout the semester. A substantial part of learning will consist of directed independent learning and the expectation is to engage with the online discussion forum about 1 h per week, as online peer support is a common feature of scientific image analysis. Coding and version control skills are assessed through checkpoints, self-reflection is assessed through feedback statements. The main assignment consists of a short pitch to non-experts to explain your approach, followed by a report aimed at your peers, which together assesses coding, communication, understanding and application to (physics) problems.
|
Entry Requirements (not applicable to Visiting Students)
| Pre-requisites |
|
Co-requisites | |
| Prohibited Combinations | |
Other requirements | None |
Information for Visiting Students
| Pre-requisites | Experimental physics or data analysis at pre-honours level (or equivalent)
Physics computer programming at JH level (or equivalent)
Fourier analysis at JH level (or equivalent)
Proficiency with Python at JH level (or equivalent) |
Course Delivery Information
| Not being delivered |
Learning Outcomes
On completion of this course, the student will be able to:
- demonstrate an understanding of scientific images, quantitative information in images, how that relates to the instrument¿s contrast mechanism, and how image analysis can be used to solve (physics) problems;
- write an image-analysis code in a contemporary programming language to extract relevant quantitative information from scientific images
- communicate image-analysis results effectively to both experts (peers) and a wider audience
- make appropriate use of version control to work collaboratively on an image-analysis code
- reflect on their own image-analysis practice and constructively identify areas for further development.
|
Reading List
- Ravishankar Chityala and Sridevi Pudipeddi, Image Processing and Acquisition using Python, CRC Press (2014)
- Sandipan Dey, Python Image Processing, Packt» Publishing (2020)
- David J. Pine, Introduction to Python for Science and Engineering, CRC Press (2019), also available (2014): https://physics.nyu.edu/pine/pymanual/html/pymanMaster.html
Python: https://www.python.org/
NumPy: http://www.numpy.org/
Matplotlib: http://matplotlib.org/
Scikit-image: https://scikit-image.org/ and https://scipy-lectures.org/packages/scikit-image/index.html
OpenCV: https://docs.opencv.org/3.4/d9/df8/tutorial_root.html
ImageJ: https://imagej.net/learn/ |
Additional Information
| Graduate Attributes and Skills |
- Generic cognitive skills: develop data skills including critically evaluating the source and quality of data/images, recognizing potential biases in the data, analyzing the data, and contextualizing the results of the analysis including potential impact.
- Communication: communicate complex information to a diverse range of audiences including peers/specialists via written reports, and a wider audience using methods appropriate for the target audience.
- Quantitative and IT skills: develop expertise in extracting quantitative information from scientific images and develop coding skills for image analysis.
- Autonomy, accountability and team work: develop image-analysis codes and/or protocols in collaboration with others, making appropriate use of version control, and being able to identify and take responsibility for a substantial contribution to the collaboration.
- Personal development: routinely reflecting critically on role and responsibilities, identify areas for development, and recognize own and team strengths.
|
| Keywords | Image analysis,data analysis,imaging,machine learning |
Contacts
| Course organiser | Dr Job Thijssen
Tel: (0131 6)50 5274
Email: j.h.j.thijssen@ed.ac.uk |
Course secretary | |
|
|
University Homepage