Search
This course covers a selection of the most used image analysis methods, with emphasis on images from medical and biological modalities, from microscopy, to ultrasound, MRI, or CT, including time sequences. We describe some of the most used algorithms for solving the key problems in this area - preprocessing, segmentation, registration, reconstruction and classification - and their use in applications. We show how to deal with the specifics of medical data such as non-linear transformation, 3D data, large variability, lack of reliable keypoints, lack of labeled training data etc. This course complements the Computer Vision Methods course, which covers techniques for images from standard optical cameras.
This course assumes the knowledge of basic image processing algorithms, as taught for example in the Digital image course. We assume the knowledge of programming, mathematics and machine learning approximately at the level of the bachelor programs here at FEE (FEL). The knowledge of medical image acquisition devices principles (taught for example in the course BEAM33ZSL) is useful but not necessary.
Lecturer: Jan Kybic (kybic@fel.cvut.cz), consultations by appointment. Alexander (Oleksandr) Shekhovtsov
Lab assistant: Denis Baručić (barucden@fel.cvut.cz), consultations by appointment.
In the lectures, we will present a number of relevant algorithms for (bio)medical image processing based on the original scientific papers. You are strongly advised to read the papers in advance. Feel free to ask questions, both in advance and during the lecture. The selection of papers can be adapted to some extent, so if there is a particular topic that you are interested in or not interested in at all, let us know. The presence at the lectures is not compulsory but is strongly encouraged.
The Labs will take place in a computer laboratory. In the first part of the semester, you will try to implement some simple algorithms, using available libraries for the more complicated methods. During the second part of the semester, you will work on your Semester work. The labs will also include ten short, graded quizzes related to the papers discussed at the lectures. If you cannot attend, let the teacher (lab assistant) know.
The Semester work should take you about 10 hours and will contribute up to 50 points towards the exam (out of 100). It will consist in an independent implementation of some biomedical image analysis algorithm or its application to a particular problem. You are encouraged to choose your own topic, so please approach us if you have an idea. Otherwise, you will be assigned a topic. You are expected to submit a short report (5-10 pages including images) and your code. You should also present your results briefly (5 minutes + equations) to your peers. The presentation (10 points) can be short but it should explain what the problem was, how others solved it, how you solved it, and how well it worked. The report (35 points) should be structured similarly but with more details, equations, images, and graphs. The code (5 points) should be commented.
The students are awarded points for the quizzes, semester work, and oral exam. The maximum number of points is 100. The following table lists the number of points for each graded unit.
To be allowed to attend the oral exam (get a “zápočet”), one needs to get at least 30 points (out of 60) from the quizzes and the semester work together.
The exam will be oral and in person (epidemiological situation permitting) and the student should demonstrate the knowledge of the basic principles of the discussed algorithms. You should know the mathematical formulation, when and why it works. In theory, given enough time, you should be able to implement the method yourself. On the other hand, it is not necessary to know all little details. It is not necessary to know the articles marked as optional. Here are a few example questions.
The minimum requirement to pass the exam is to earn 20 points (out of 40). Assuming this requirement is met, the final grade will be determined based on the sum of all points earned according to the following scale.
This course allows the use of artificial intelligence (AI) tools such as ChatGPT, Google Bard, Github Copilot, and others. However, any AI-generated output (or anything based on AI-generated output) must be clearly denoted and acknowledged in any submitted work. Tell us (in at least a paragraph or more) where you used AI and how and whether it (in your opinion) improved the quality of your work and your learning. Using AI without acknowledging it will be considered as cheating. When in doubt, write what you did. Spell-checkers are not considered AI but anything more powerful should be declared. Failure to properly acknowledge the sources of any submitted work will be considered plagiarism. You are in any case fully responsible for the result (both the text and code). Students must understand and be able to explain all parts of their submitted work.
The goal of the course is for students to learn and this goal must be respected when employing any AI tools. Consider using AI as a feedback generator or a personal tutor.