Click on image to get a larger view!
This page is outdated; for an update version and course materials, please click here!

Course Outline

The course gives an overview of shape modeling and its use in image analysis. Using theory developed in the field of statistics and machine learning, we will discuss a unifying framework for shape modeling using Gaussian processes, and show the connectsion to many popular methods in image analysis and shape modeling.

VVNr.

3952501

Lecturer

Marcel Lüthi, Marcel.Luethi@unibas.ch

Assistant/Exercises

Ghazi Bouabene, Ghazi.Bouabene@unibas.ch

Time, Location

Wednesday, 08.15  10.00, Spiegelgasse 1, Seminarraum 03
Start: 18. 2. 2015 (weekly lectures)

Exercises

Practical programming exercises. The students will implement the basic building blocks of a model based image analysis system, building up on an
existing framework, developed at the University of Basel. The exercises will be done in the programming language Scala, but no prior knowledge of Scala is required.
Regular counselling times will be announced at the beginning of the course.

Examination

The examination consists of a project and a written report.
For the project the students will need to implement a system for modelbased segmentation of an anatomical structure. It will be organized as a competition, where the performance will be evaluated on a publicly available benchmark dataset.
The project will be done in small groups. A report, which summarizes the research findings in the project, will need to be handed in individually.

Grading

The grading will be done based on the performance on the practical project and the scientific quality of the report. Grades are on the standard scale from 1 to 6.

Prerequisites

 Linear algebra
 Probability theory and Statistics.
 Basic programming skills.

Goal

At the end of the course the students will:
 know the theory of Gaussian processes and how they can be applied for shape modeling.
 understand how methods from machine learning can be used to learn properties of shapes.
 be able to implement a basic system for image analysis and shape modeling
 have a good overview of methods used for shape modeling.
 have a good overview of different methods for image registration and can see the connections between different classes of algorithms

Content

The guiding topic through the course is the problem of model based image analysis. We start by discussing some basic approaches for image analysis and in particular
focus on "Analysis by synthesis". This approach is based on the assumption that in order to be able to analyse an image, one also needs to be able to synthesize this image by mean of probabilistic models. The main topic of this course will be how we can build such models, and use them for analysing (medical) images and the reconstruction of partial data.
The theoretical basis for our method is the theory of Gaussian processes, which are heavily used in the fields of statistics and machine learning. A major part of the lectures will be to study Gaussian processes and their application for modelling shapes. We will take in this course a machine learning perspective, where the topic of modeling with Gaussian processes has been explored in depth and many algorithms and theoretical results have been established. By applying the methods for shape modeling, the concepts can be visualized and thus it will be easier to obtain a good intuition of these methods. This aspect will in particular be covered in the exercises, where the theoretical methods are implemented to build a complete framework for modelbased image analysis.
The theory of Gaussian processes and Reproducing Kernel Hilbert spaces is very general and can be used to unify many distinct methods currently used in image analysis.
We aim at making wherever possible the connections to popular research approaches in image analysis, shape modelling and registration, such that a student gets a good overview and understanding of the methods used in this field.
