The lectures are given in building 321, room 033, on Thursdays from 1 pm – 2.30 pm (approximately). The computer exercises take place immediately after the lectures in the same room. Terminals are reserved until 5 pm; a teaching assistant will be present. You may also use the terminals outside this time-window, provided you can find a terminal that is not occupied. The computer exercises are on the topic of last week’s lecture. The computer exercises should be carried out in groups of 2-3.
Lecture notes and exercise texts and materials are available through CampusNet (one week in advance - as a rule).
For each subject considered in the course an exercise is carried out and an exercise report must be prepared and handed in. The course examination consists of an oral examination in the course topics based on the course theory and exercise reports. The oral examination will be held during the period 9-21 December 2013 (the exact date will be announced later).
This is a 5 ECTS point course corresponding to a work load of 8-10 hours work per week and most of this time is used to implement the image analysis models in the exercises. Some of the algorithms are complicated and the computations can be time-consuming. You can only understand these algorithms fully by implementing them, so that you can test the effect of varying data and parameters. Hence a successful completion of the course (and exam) requires that your implementation works and that you demonstrate a good understanding of the strengths and weaknesses of the models.
Introduction; Fitting of image functions
Linear basis function models;
maximum likelihood estimation and least squares;
regularization; image smoothing;
|2||12 Sep||kvle||Landmark based registration||MR bias field correction|
Intensity based methods for registration
Sum-of-squared differences; correlation;
mutual information; principal axis transform
|4||26 Sep||Tron Darvann||EXCURSION: 3D laboratory. School of Dentistry||No exercise|
|5||3 Oct||kvle||Non-linear deformations||Mutual information based registration|
|6||10 Oct||kvle||Surface based registration||non-linear intensity-based registration|
|7||24 Oct||Kristoffer Hougaard Madsen||EXCURSION: Danish Research Center for Magnetic Resonance (DRCMR)||No exercise|
Active contour model; dynamic programming
|non-linear intensity-based registration (cont.)|
Voxel-based segmentation I
Gaussian mixture model; Markov random field priors; mean-field approximation
Voxel-based segmentation II
Expectation-Maximization algorithm; MR bias field correction
|11||21 Nov||Mads Nielsen, Biomediq||Quantification of medical images for clinical trials||Brain tumor segmentation|
Reference templates; group-wise registration; probabilistic atlases; label propagation
|13||5 Dec||kvle||Validation of segmentation methods||Exercise catch-up|